Image quality enhancement circuit and method therefor

The present invention relates to image quality enhancement, and more particularly, to an image quality enhancing circuit having functions such as noise reduction, contrast enhancement based on histogram equalization, local contrast enhancement and colour compensation and a method therefor.

In general, the image quality of a video signal can be deteriorated due to various factors. A low contrast can be a factor among representative ones of image quality degradation. As a method for correcting the image quality degradation, there are gamma correction for making a correction according to a variation in brightness, histogram equalization, etc.

The principal operation of the histogram equalization is to convert a given input image on the basis of the histogram of the input image. Here, the histogram represents gray level distribution at a given input image. Such a histogram of the gray level provides an entire depiction on the appearance of an image. A gray level appropriately adjusted according to sample distribution of an image enhances the appearance and contrast of the image.

The histogram equalization for enhancing the contrast of a given image according to sample distribution of the image is the most widely known among various contrast enhancing methods, which is disclosed in the following documents: [1] J.S.Lim, "Two-Dimensional Signal and Image Processing," Prentice Hall, Englewood Cliffs, New Jersey, 1990; and [2] R.C.Gonzalez and P.Wints, "Digital Image Processing," Addison-Wesley, Reading, Massachusetts, 1977.

Also, useful applications of the histogram equalization including medical image processing and radar image processing are disclosed in the following documents: [3] J.Zimmerman, S.Pizer, E.Staab, E.Perry, W.McCartney, and B.Brenton, "Evaluation of the effectiveness of adaptive histogram equalization for contrast enhancement," IEEE Tr.on Medical Imaging, pp. 304-312, Dec. 1998; and [4] Y.Li, W.Wang, and D.Y.Yu, "Application of adaptive histogram equalization to x-ray chest image," Proc. of the SPIE, pp. 513-514, vol. 2321,1994. Accordingly, a technique using the histogram of a given image has been usefully applied to various fields such as the medical image processing, infrared ray image processing, and radar image processing.

In general, since histogram equalization has the effect of stretching a dynamic range, it flattens distribution density of a resultant image. Therefore, the contrast of the image is enhanced. Such a well-known characteristic of the histogram equalization becomes a defect in some actual cases. That is, because of this characteristic of the histogram equalization flattening an output density, the average brightness of the output image approaches to a middle gray level. In practice, for histogram equalization of an analog image, the average brightness of an output image in the histogram equalization is exactly the middle gray level regardless of the average brightness of an input image. Obviously, the above-described characteristic is not desirable for the practical application. For example, a problem that a scene photographed at night looks too bright after the histogram equalization is generated.

As another factor of the image quality degradation, there is an impulse noise. The independent impulse noise is uniformly distributed on a frequency region. As a result, application of a simple linear filter causes details of an image to blur and a high frequency component of the impulse noise not to be effectively removed.

Still another factor of the image quality degradation is that the gamma correction or the histogram equalization for enhancing a low contrast enhances an entire contrast of a video signal but is not so effective for enhancement of a contrast at detailed portions being visually more important information, i.e., a local contrast.

Yet another factor of the image quality degradation is that, unless colour compensation is performed on a colour signal according to a variation in luminance occurring when a predetermined luminance processing such as the histogram equalization is performed on a luminance signal to enhance a contrast, a primary colour signal is distorted.

Accordingly, it is an aim of embodiments of the present invention to provide an image quality enhancing circuit having such functions as noise reduction, contrast enhancement based on mean-separation histogram equalization, local contrast enhancement, and colour compensation.

It is another aim of embodiments of the present invention to provide an image quality enhancing method having such functions as noise reduction, contrast enhancement based on mean-separation histogram equalization, local contrast enhancement, and colour compensation.

The present invention provides, a noise reducing means of an image quality enhancing circuit detects an impulse for each of an input luminance signal and input colour signals and outputs a trimmed mean of a predetermined size window when the impulse is detected, and otherwise, bypasses the input signals without change. A contrast enhancing means independently equalizes a luminance signal of a picture unit output by the noise reducer by obtaining a histogram of subimages divided on the basis of the mean value of the luminance signal and outputs an enhanced luminance signal. A local contrast enhancing means detects a local contrast defined as a difference between the value of an input sample with respect to the enhanced luminance signal and each value obtained by low-pass filtering samples in a predetermined size window including the input sample, adaptively weights the input sample value according to the detected local contrast, and outputs a changed luminance signal. Also, a colour compensating means compensates the colour signals output by the noise reducing means according to the changed luminance signal.

According to an aspect of the invention, there is provided an image quality enhancing circuit for enhancing image quality through a predetermined image processing on an input image signal, said circuit comprising: noise reducing means for detecting an impulse for each of an input luminance signal and input colour signals and outputting a trimmed mean of a predetermined size window when said impulse is detected, and otherwise, bypassing said input signals without change; contrast enhancing means for equalizing a luminance signal of a picture unit output by said noise reducing means and outputting an enhanced luminance signal; local contrast enhancing means for detecting a local contrast defined as a difference between the value of an input sample with respect to said enhanced luminance signal and each value obtained by low-pass filtering samples in a predetermined size window including the input sample, adaptively weighting said input sample value according to said detected local contrast, and outputting a changed luminance signal; and colour compensating means for compensating said colour signals output by said noise reducing means according to said changed luminance signal.

Preferably, the image quality enhancing circuit further comprises:    a first colour converter for converting an input RGB signal into a luminance signal and colour signals and outputting the result to said noise reducing means; and a second colour converter for outputting said luminance signal output by said local contrast enhancing means and said compensated colour signal output by said colour compensating means in RGB form.

Said first colour converter preferably converts said input RGB signal into one of signals (Y, U, V) and (Y, I, Q).

Said noise reducing means preferably comprises a plurality of noise reducers for reducing impulse noise for said luminance signal and said colour signals, respectively.

Said noise reducing means may comprise: a first noise reducer for reducing impulse noise included in the input luminance signal; a selector for selecting a signal from two input colour signals; a second noise reducer for reducing impulse noise included in the colour signal selected by said selector; and a separator for separating the colour signal whose noise is reduced by said second noise reducer into two types of colour signals.

Each of said noise reducers may comprise: a window generator for generating a plurality of different-sized windows including an input signal; a plurality of outlier value detectors for detecting a mean between a mean sample value for each window and an absolute deviation of samples and detecting whether an input signal has an outlier value; a selection control signal generator for generating a selection control signal when all of said input signals are detected to be outlier values in said plurality of outlier value detectors; a trimmed mean filter for trimming an input signal using a trimming window of a predetermined size and outputting a trimmed mean; and a selector for outputting said trimmed mean as an output signal when an impulse component exists in said input signal according to said selection control signal and, otherwise, outputting said input signal as an output signal without change.

Said window generator may substitute said input signal for an output signal in a recursive noise reduction mode according to a recursive/nonrecursive noise reduction mode signal and then generates a plurality of windows for a next input signal.

Said window generator may generate first and second windows each having a different size and including the input signal.

Said plurality of outlier value detectors preferably comprise: a first outlier value detector for judging an impulse component to be included in an input signal when an absolute difference between the input signal and a mean sample value of said first window is greater than the mean of an absolute deviation of said first window multiplied by a parameter for predetermined noise reduction, and outputting a first outlier value detecting signal; and a second outlier value detector for judging an impulse component to be included in an input signal when an absolute difference between the input signal and a mean sample value of said second window is greater than the mean of an absolute deviation of said second window multiplied by the parameter for predetermined noise reduction, and outputting a second outlier value detecting signal.

Said trimming window preferably uses one window among said plurality of windows.

Said contrast enhancing means is preferably comprised of a mean-separate histogram equalizer for independently equalizing a luminance signal of a picture unit output by said noise reducing means by obtaining a histogram of subimages divided on the basis of the mean value of said luminance signal.

Said contrast enhancing means may comprise: first calculating means for receiving a noise-removed luminance signal output by said noise reducing means, in a picture unit, and calculating the distribution of a gray level; second calculating means for receiving said noise-reduced luminance signal in a picture unit and calculating a mean level; third calculating means for dividing said gray level distribution of the calculated picture unit into a predetermined number of subimages according to said mean level and calculating a cumulative density function for each subimage; and mapping means for mapping said noise-reduced luminance signal to a gray level according to said cumulative density function value calculated by each of said subimages and outputting an enhanced luminance signal.

Said picture unit is preferably a frame unit and said predetermined number is two.

Preferably, a frame memory is provided for delaying said noise-reduced luminance signal in a frame unit in order to input a signal of the same frame as said cumulative density function calculated by said third calculating means to said mapping means.

A buffer may be provided for renewing said cumulative density function calculated by said third calculating means in a picture unit and providing the cumulative density function value stored during the renewal to said mapping means.

Said mapping means may comprise: a first mapper for mapping said noise-reduced luminance signal to a gray level having a first range according to a corresponding cumulative density function value when said noise-reduced luminance signal is a first subimage lower than or equal to the mean level; a second mapper for mapping said noise-reduced luminance signal to a gray level having a second range according to a corresponding cumulative density function value when said noise-reduced luminance signal is a second subimage greater than said mean level; a comparator for comparing said noise-reduced luminance signal with said mean level and generating a selection control signal; and a selector for selecting said first mapper when said noise-reduced luminance signal is said first subimage according to said selection control signal, and otherwise, selecting said second mapper.

Said mapping means may comprise: a first mapper for mapping the luminance signal output by said frame memory to a gray level having a first range according to a corresponding cumulative density function value when said luminance signal output by said frame memory is a first subimage lower than or equal to the mean level; a second mapper for mapping said luminance signal output by said frame memory to a gray level having a second range according to a corresponding cumulative density function value when said luminance signal output by said frame memory is a second subimage greater than said mean level; a comparator for comparing said luminance signal output by said frame memory with said mean level and generating a selection control signal; and a selector for selecting said first mapper when said luminance signal output by said frame memory is said first subimage according to said selection control signal, and otherwise, selecting said second mapper.

Said contrast enhancing means preferably further comprises brightness compensating means for outputting a compensated mean level by adding a corrected value depending on a predetermined correction function to said mean level.

Said brightness compensating means preferably outputs a compensated mean level by adding a corrected value greater than zero to said mean level when said mean level is very small and by adding a corrected value lower than zero when said mean level is very large.

Said mapping means may comprise: a first mapper for mapping said noise-reduced luminance signal to a gray level from a minimum gray level to a compensated mean level (B_m) according to the cumulative density function value of a corresponding subimage when said noise-reduced luminance signal is lower than or equal to said mean level; a second mapper for mapping said luminance signal to a gray level from a changed compensated mean level (B_m') to a maximum gray level (X_L-1) according to the cumulative density function value of a corresponding subimage when said luminance signal is greater than said mean level, wherein B_m' equals B_m+X_L-1/(L-1); a comparator for comparing said noise-reduced luminance signal with said mean level and generating a selection control signal; and a selector for selecting said first mapper when said noise-reduced luminance signal is said first subimage according to said selection control signal, and otherwise, selecting said second mapper.

Said contrast enhancing means may further comprise gain control means for controlling the gain of said enhanced luminance signal according to an amount of variation in gray level between said noise-reduced luminance signal and said enhanced luminance signal and the level of said noise-reduced luminance signal.

Said gain control means may control the gain of said enhanced luminance signal on the basis of a Weber ratio.

Said gain control means may comprise: a subtracter for subtracting said noise-reduced luminance signal from said enhanced luminance signal and detecting a variation amount corresponding to the difference; a gain feature determiner for calculating a maximum bounding function value according to the level of said noise-reduced luminance signal to restrict enhancement of said noise-reduced luminance signal using a predetermined maximum bounding function; a multiplier for multiplying said maximum bounding function value by a gain control parameter having a predetermined value and outputting a bounding function value; a limiter for comparing said bounding function value with said variation amount, limiting said variation amount according to the compared result, and outputting a limited variation amount; and an adder for adding said limited variation amount to said noise-reduced luminance signal.

Said gain control means may further comprise a selector arranged to compare said noise-reduced luminance signal with said mean level and select the gain control parameter to be a first gain control parameter when said noise-reduced luminance signal is lower than or equal to said mean level and, otherwise, to select the gain control parameter to be a second gain control parameter.

Said maximum bounding function g(X_k) is preferably aX_k and a may be a constant or said maximum bounding function g(X_k) may be <maths id="math0001"><math display="inline"><mrow><mtext>a</mtext><msqrt><msub><mrow><mtext mathvariant="italic">X</mtext></mrow><mrow><mtext mathvariant="italic">k</mtext></mrow></msub></msqrt></mrow></math><img file="EP0833501A2_D0001.tif" /></maths> where a is a constant.

Preferably, said limiter outputs said variation amount as said limited variation amount when an absolute value of said variation amount is lower than or equal to said bounding function value, and otherwise, outputs said bounding function value as said limited variation amount.

Said local contrast enhancing means may comprise: a window generator for generating an input sample with respect to an enhanced luminance signal output by said contrast enhancing means generating a predetermined-sized window including said input sample; a low-pass filter for low-pass filtering the samples of said window generated by said window generator and outputting a low-pass filtered signal; a subtracter for subtracting said low-pass filtered signal from the input sample output by said window generator and generating a local contrast signal corresponding to a contrast which can be visually perceived by a human being; an absolute value calculator for calculating an absolute value of said local contrast signal and detecting a region having high or low contrast; a weighting unit for adding a weighted value obtained by a predetermined weighting function to the output of said absolute value calculator; a first multiplier for multiplying the output of said weighter by a parameter for controlling an amount of local contrast enhancement; a second multiplier for multiplying said local contrast signal by the output of said first multiplier; and an adder for adding the output of said second multiplier to the input sample output by said window generator and outputting a changed luminance signal.

Said weighting function preferably enhances said local contrast signal when the absolute value of said local contrast signal is lower than a predetermined value and bypasses said local contrast signal without enhancement when the absolute value of said local contrast signal is greater than or equal to the predetermined value.

Said colour compensating means may comprise: an operator for operating a ratio between the luminance signal output by said noise reducing means and the changed luminance signal output by said local contrast enhancing means; and multipliers for multiplying the colour signals output by said noise reducing means by the ratio output by said operator and outputting compensated colour signals.

Said colour compensating means may comprise: an operator for operating a ratio between the luminance signal output by said noise reducing means and the changed luminance signal output by said local contrast enhancing means; and regulators for receiving the colour signals output by said noise reducing means, the ratio output by said operator, and a predetermined parameter and outputting compensated colour signals by changing the value of the colour signals using a predetermined compensating line.

The ratio between the luminance signal output by said noise reducing means and the changed luminance signal output by said local contrast enhancing means is given as q, said compensating line may be expressed by following equation:<maths id="math0002"><math display="block"><mrow><mtext>(R',G',B')=q(R,G,B).</mtext></mrow></math><img file="EP0833501A2_D0002.tif" /></maths>

When the ratio between the luminance signal output by said noise reducing means and the changed luminance signal output by said local contrast enhancing means is given as q, said compensating line may be expressed by following equation:<maths id="math0003"><math display="block"><mrow><mtext>(R',G',B')=A(R,G,B)+K,</mtext></mrow></math><img file="EP0833501A2_D0003.tif" /></maths> wherein A is <maths id="math0004"><math display="inline"><mrow><mfrac><mrow><mtext mathvariant="italic">q</mtext><mtext>(1-α),</mtext></mrow><mrow><mtext mathvariant="italic">q</mtext><mtext>-1</mtext></mrow></mfrac></mrow></math><img file="EP0833501A2_D0004.tif" /></maths> , K is <maths id="math0005"><math display="inline"><mrow><mfrac><mrow><mtext>(</mtext><mtext mathvariant="italic">q</mtext><mtext>-1)α</mtext></mrow><mrow><mtext mathvariant="italic">q</mtext><mtext>-1</mtext></mrow></mfrac></mrow></math><img file="EP0833501A2_D0005.tif" /></maths><i>Max</i>, α is greater than or equal to 0 and smaller than or equal to 1, α is a predetermined parameter, and Max is a maximum value.

According to a further aspect of the invention there is provided an image quality enhancing method for enhancing the image quality of an input image signal by performing a predetermined image processing on said input image signal, said method comprising the steps of: (a) detecting an impulse for each of input luminance and colour signals, outputting a trimmed mean of a predetermined- sized window when said impulse is detected, and bypassing said input signals when said impulse is not detected, thereby outputting noise-reduced luminance and colour signals; (b) receiving said noise-reduced luminance signal in a picture unit, equalizing said noise-reduced luminance signal, and outputting an enhanced luminance signal; (c) detecting a local contrast defined as a difference between an input sample value with respect to said enhanced luminance signal and each value obtained by low-pass filtering samples of a predetermined-sized window including the input sample, and outputting a luminance signal changed by adaptively weighting the input sample value according to said detected local contrast; and (d) compensating said noise-reduced colour signals according to said changed luminance signal and outputting compensated colour signals.

The method preferably further comprises the steps of: (el) converting input RGB signals into a luminance signal and colour signals and outputting the result before said step (a) ; and (e2) converting and outputting said changed luminance signal and said compensated colour signal in RGB signal form after said step (d).

Said RGB signal received in step (el) is preferably converted into one of signals (Y, U, V), (Y, R-Y, B-Y) and (Y, I, Q).

Said step (a) may comprise the substeps of: (al) generating a plurality of windows of different sizes including the input signal; (a2) obtaining a mean sample value for each of said windows and a mean of an absolute deviation between samples; (a3) detecting whether or not an impulse component exists in the input signal, using said obtained mean sample value of each window and said means of the absolute deviation between samples; and (a4) outputting a trimmed mean by trimming the samples in a predetermined-sized trimming window when an impulse component is detected from said input signal, and otherwise, bypassing said input signal.

When said impulse component is detected from said input signal in at least two windows in step (a3), it is preferably judged that said impulse component exists in said input signal.

Said trimming window may be one of said plurality of windows.

Said step (a) may further comprise the step of (a5) replacing said input signal with said output signal, moving it to a next input signal, and returning to step (a1).

Said step (b) preferably comprises the substeps of: (bl) receiving said noise-reduced luminance signal in a picture unit and calculating a mean level; (b2) calculating a cumulative density function, based on gray level distribution, for every subimage divided on the basis of said calculated mean level; and (b3) outputting an enhanced luminance signal by performing independent histogram equalization with respect to each subimage on the basis of said cumulative density function value obtained for each of said subimages.

Said noise-reduced luminance signal of a picture unit is preferably divided into two subimages according to said mean level, in step (b2).

Said step (b3) may comprise the substeps of: (b31) mapping the samples for each subimage to a gray level according to said cumulative density function value obtained for every subimage; and (b32) comparing the level of said noise-reduced luminance signal with said mean level and selecting one from said signals mapped to the gray level in each of said subimages, according to a compared result.

The method may further comprise the step of (b33) delaying said noise-reduced luminance signal in a picture unit and outputting said delayed signal to step (b32).

Said step (b) may further comprise the step of (b4) outputting a compensated mean level by adding a corrected value depending on a predetermined correcting function to said mean level. In step (b4), said compensated mean level is preferably output by adding a corrected value greater than zero to said mean level when said mean level is very small, and by adding a corrected value lower than zero to said mean level when said mean level is very large.

Said step (b3) may comprise the substeps of: (b31') mapping said noise-reduced luminance signal to a gray level from a minimum gray level to a compensated mean level (B_m) according to the cumulative density function value of a corresponding subimage when said noise-reduced luminance signal is lower than or equal to said mean level; and (b32') mapping said luminance signal to a gray level from a changed compensated mean level (B_m') to a maximum gray level (X_L-1) according to the accumulative density function value of a corresponding subimage when said luminance signal is greater than said mean level, wherein B_m' equals B_m+X_L-1/(L-1).

Said step (b) may comprise the step of (b5) controlling the gain of an enhanced signal according to an amount of gray level variation between said noise-reduced luminance signal and said enhanced luminance signal and the level of said noise-reduced luminance signal. Said gain control of said enhanced luminance signal is preferably based on a Weber ratio, in step (b5).

Said step (b5) may comprise the substeps of: (b51) subtracting said luminance signal from said enhanced signal and detecting an amount of variation corresponding to the difference; (b52) calculating a limit enhancement amount of said luminance signal using a predetermined maximum bounding function and a predetermined gain control parameter; (b53) comparing said limit enhancement amount with said variation amount, limiting said variation amount according to the compared result, and outputting a limited variation amount; and (b54) adding said limited variation amount to said noise-reduced luminance signal. Said step (b52) may comprise the substeps of: (b521) calculating a maximum bounding function value according to the level of said noise-reduced luminance signal using a predetermined maximum bounding function in order to limit enhancement of said noise-reduced luminance signal; and (b522) multiplying said maximum bounding function value by a predetermined gain control parameter and outputting a limit enhancement amount.

Said predetermined gain control parameter may comprise a plurality of gain control parameters for each divided subimage.

Said predetermined gain control parameter may comprise a gain control parameter which is commonly applied to said noise-reduced luminance signal.

Said maximum bounding function g(X_k) may be aX_k where a is a constant or said maximum bounding function g(X_k) may be <maths id="math0006"><math display="inline"><mrow><mtext>a </mtext><msqrt><msub><mrow><mtext mathvariant="italic">X</mtext></mrow><mrow><mtext mathvariant="italic">k</mtext></mrow></msub></msqrt></mrow></math><img file="EP0833501A2_D0006.tif" /></maths> and a is a constant.

In step (b53), said variation amount is preferably output as said limited variation amount when an absolute value of said variation amount is lower than or equal to said limit enhancement amount, and otherwise, said limit enhancement amount is output as said limited variation amount.

Said step (c) preferably comprises the substeps of: (cl) detecting a local contrast corresponding to a difference between an input sample value with respect to said enhanced luminance signal and each value obtained by low-pass filtering the samples in a predetermined-sized window including an input sample; and (c2) adaptively weighting a weighted value depending on a predetermined weighting function to said input sample value according to said detected local contrast value.

Said weighting function preferably enhances said local contrast value when said detected local contrast value is lower than a predetermined value and bypasses said local contrast value without enhancement when said detected local contrast value is greater than or equal to the predetermined value.

Said compensated colour signal may be changed in the same direction as that of said noise-reduced colour signal, in step (d).

In said step (d), said changed luminance signal preferably forms a luminance plane changed in a colour signal space, and luminance values of every colour signal in said formed luminance plane are the same.

In step (d), said compensated colour signal may be obtained by an intersecting point of a line linking the direction of said noise-reduced colour signal with the changed luminance plane.

When a ratio between said noise-reduced luminance signal and said changed luminance signal is given as q, said compensating line may be expressed by following equation:<maths id="math0007"><math display="block"><mrow><mtext>(R',G',B')=q(R,G,B).</mtext></mrow></math><img file="EP0833501A2_D0007.tif" /></maths>

When a ratio between said noise-reduced luminance signal and said changed luminance signal is given as q, said compensating line may be expressed by following equation:<maths id="math0008"><math display="block"><mrow><mtext>(R',G',B')=A(R,G,B)+K,</mtext></mrow></math><img file="EP0833501A2_D0008.tif" /></maths> wherein A is <maths id="math0009"><math display="inline"><mrow><mfrac><mrow><mtext mathvariant="italic">q</mtext><mtext>(1-α)</mtext></mrow><mrow><mtext mathvariant="italic">q</mtext><mtext>-1</mtext></mrow></mfrac></mrow></math><img file="EP0833501A2_D0009.tif" /></maths>, K is <maths id="math0010"><math display="inline"><mrow><mfrac><mrow><mtext>(</mtext><mtext mathvariant="italic">q</mtext><mtext>-1)α</mtext></mrow><mrow><mtext mathvariant="italic">q</mtext><mtext>-1</mtext></mrow></mfrac></mrow></math><img file="EP0833501A2_D0010.tif" /></maths><i>Max</i>, α is greater than or equal to 0 and lower than or equal to 1, α is a predetermined parameter, and Max is a maximum value.

For a better understanding of the invention, and to show how embodiments of the same may be carried into effect, reference will now be made, by way of example, to the accompanying diagrammatic drawings, in which: <ul id="ul0001" list-style="none"><li>Figure 1 is a block diagram of an image quality enhancing circuit according to an embodiment of the present invention;</li><li>Figure 2 is a detailed block diagram of the first noise reducer shown in Figure 1;</li><li>Figure 3 is a detailed circuit diagram of the window generator shown in Figure 2;</li><li>Figure 4 is a detailed block diagram of the contrast enhancer based on mean-separation histogram equalization having gain control and brightness compensation, shown in Figure 1;</li><li>Figures 5a and 5b show examples of brightness correcting functions which are applied to the brightness compensator shown in Figure 4;</li><li>Figures 6a and 6b show the relationships between average levels compensated by the brightness correcting functions shown in Figures 5a and 5b and those of input images, respectively;</li><li>Figure 7 is a detailed block diagram of the local contrast enhancer shown in Figure 1;</li><li>Figure 8 is a detailed circuit diagram of the MXN window generator shown in Figure 7;</li><li>Figure 9 shows the characteristic of a weighting function applied to a weighting unit shown in Figure 7;</li><li>Figure 10 is a view for illustrating colour compensation according to a luminance variation;</li><li>Figure 11 shows a colour compensation line for preventing colour saturation;</li><li>Figure 12 is a detailed circuit diagram of the colour compensator shown in Figure 1 according to an embodiment of the present invention; and</li><li>Figure 13 is a detailed circuit diagram of the colour compensator shown in Figure 1 according to another embodiment of the present invention.</li></ul>

Referring to Figure 1, a first colour converter 100 receives baseband digital colour signals indicated by R_in, G_in and B_in and converts them into a digital luminance signal and system-defined colour signals which are indicated by Y_in, U_in and V_in.

Here, R, G and B signals can be converted into various different colour signals like (Y, I, Q), (Y, U, V) and (Y, R-Y, B-Y) colour systems according to a signal processing method. The relationship between the R, G and B signals and new colour signals Y, U and V can be expressed by following equation 1.<maths id="math0011"><img file="EP0833501A2_D0011.tif" /></maths>

In embodiments of the present invention, (Y,U,V) is defined by a colour system, Y denotes a luminance signal, and U and V represent system-defined colour signals.

A first function of an image quality enhancing circuit is to remove impulse noise. First to third noise reducers 200, 220 and 240 detect an outlier value based on statistics of samples and trim the result in order to remove impulse noise, respectively. In other words, the first to third noise reducers 200, 220 and 240 output the trimmed mean of a given window when an impulse has been detected from an input sample, otherwise, bypass the input sample.

Also, statistics of samples calculated by two windows of different sizes is useful to reduce detection failure of the impulse. The detection failure can be an essential factor of blurring out details of a video signal. The degree of detection of an outlier value is determined by a parameter k.

Accordingly, noise reducers 200, 220 and 240 having independent but identical configurations are employed to reduce impulse noises of the Y_in, U_in, and V_in signals.

However, since a typical digital standard image has a 4:2:0 format or a 4:2:2 format, a single noise reducer can be employed instead of the second and third noise reducers 220 and 240 for U_in and V_in signals. In order to employ a single noise reduction block, a multiplexer for selecting the U_in or V_in signals can be installed in front of the noise reduction block, and a demultiplexer for separating noise-reduced U_in or V_in signal can be provided in rear of the noise reduction block.

A main function of the image quality enhancing circuit is to enhance a contrast using a contrast enhancer 300 based on mean-separate histogram equalization having gain control and brightness compensation (hereinafter, referred to as a contrast enhancer).

According to the basic concept of mean-separate histogram equalization as proposed by the present invention, a given image is divided into two individual groups on the basis of the mean value of the given image, and the divided sub-images are independently equalized.

If the proposed mean-separate histogram equalization is applied, an abrupt change in brightness and artifects, which can be generated after a general histogram equalization when an input image has a concentrated distributed histogram, can be effectively prevented. Also, the functions of gain control and brightness compensation can combine with the mean-separate histogram equalization. However, other embodiments of histogram equalization as well as the proposed mean-separate histogram equalization can be applied in the present invention.

The brightness compensation can be simply carried out by mapping a current mean to a desired output mean during the mean-separate histogram equalization. Also, the gain control functions to prevent excess contrast enhancement by controlling the gain of a signal enhanced by the mean-separate histogram equalization depending on the level of an input signal.

A local contrast enhancer 400 for sharpness enhancement detects a local contrast which is defined as a difference between an input sample value and a respective low-pass filtered sample values within a predetermined-size window including the input sample, and adaptively weights the input sample by applying a weight value depending on a weighting function to the local contrast according to the amplitude of the detected local contrast.

A colour compensator 500 appropriately compensates for colour signals U and V output by the second and third noise reducers 220 and 240, according to a luminance signal Y output by the first noise reducer 200 and a changed luminance signal Y' output by the local contrast enhancer 400. The colour compensator 500 maps a current colour signal put on a given luminance plane to a colour signal obtained by moving in a colour direction until the colour signal intersects an enhanced luminance plane in a RGB space. Consequently, compensated colour signals U' and V' are output by changing the colour signals U and V by the same ratio as that of the change of the luminance signal.

A second colour converter 600 receives the signal Y' output by the local contrast enhancer 400 and the U' and V' signals output by the colour compensator 500 and outputs R_out, G_out and B_out signals through an inverse processing of the conversion performed by the first colour converter 100.

The external parameters shown in Figure 1 are useful and characterize the entire function of the image quality enhancing circuit, which are a parameter (k) for noise reduction degree, a parameter (S) for selection of a recursive/nonrecursive noise reduction mode, gain control parameters (γ_u, γ_L) which are input to the contrast enhancer 300, a sharpness control parameter (β) which is input to the local contrast enhancer 400 and a colour control parameter (α) which is input to the colour compensator 500.

Figure 2 is a detailed block diagram of the first noise reducer 200 shown in Figure 1. Here, although the second and third noise reducers 220 and 240 have the same configurations, only the first noise reducer 200 for removing impulse noise of a Y_in signal will be described.

Referring to Figure 2, a window generator 202 outputs two windows expressed by the following equations, and a current input sample Y[i][j].<maths id="math0012"><img file="EP0833501A2_D0012.tif" /></maths><maths id="math0013"><img file="EP0833501A2_D0013.tif" /></maths> wherein, W_L and W_M are called as large and small windows, respectively. That is, L1 is greater than M1. The composition samples of the large and small windows output by the window generator 202 vary with a recursive/nonrecursive noise reduction mode signal (S). For this explanation, an example of L₁=L₂=2 and M₁=M₂=1 will be taken.

That is, in case of a nonrecursive noise reduction mode, 5×5 large window (W_L) and 3×3 small window (W_M) generated by the window generator 202 are as follows.<maths id="math0014"><img file="EP0833501A2_D0014.tif" /></maths><maths id="math0015"><img file="EP0833501A2_D0015.tif" /></maths>

Also, in case of a recursive noise reduction mode, 5X5 large window (W_L) and 3×3 small window (W_M) generated by the window generator 202 are as follows.<maths id="math0016"><img file="EP0833501A2_D0016.tif" /></maths><maths id="math0017"><img file="EP0833501A2_D0017.tif" /></maths>

Here, Y_N[ ][ ] denotes a signal output by the noise reducer 200 shown in Figure 2 according to the present invention, i.e., a filtered signal.

As shown in Figure 3, the detailed circuit diagram of the window generator 202 is comprised of a plurality of sample delays 261 to 280, four line delays 281 to 284 and a multiplexer 285. Here, the multiplexer 285 selects an input sample Y[i][j] upon the nonrecursive noise reduction mode and selects a final output signal Y_N[i][j] output by a selector 216 upon the recursive noise reduction mode, according to the external recursive/nonrecursive noise reduction mode signal (S). Also, D and H denote a sample delay and a line memory, respectively.

A first mean and deviation calculator 204 receives samples of the large window (W_L) generated by the window generator 202, obtains a mean sample value (A_L) of the large window (W_L) using the following equation 8, and calculates the mean (D_L) of the absolute deviations of the samples of the large window (W_L) using the following equation 9.<maths id="math0018"><img file="EP0833501A2_D0018.tif" /></maths><maths id="math0019"><img file="EP0833501A2_D0019.tif" /></maths>

A first outlier value detector 206 outputs a first outlier value detecting signal by determining that an impulse component is included in the input sample Y[i][j], when an absolute difference between the input sample Y[i][j] and the mean sample value (A_L) of the large window (W_L) is greater than the mean (kD_L) of an absolute deviation multiplied by a predetermined parameter (k), i.e.,| Y[i][j]-A_L | &gt; kD_L. Here, k denotes a noise reduction parameter.

A second mean and deviation calculator 208 receives samples of the small window (W_M) generated by the window generator 202, obtains a mean sample value (A_M) of the small window (W_M) using the following equation 10, and calculates the mean (D_M) of the absolute deviations of the samples of the small window (W_M) using the following equation 11.<maths id="math0020"><img file="EP0833501A2_D0020.tif" /></maths><maths id="math0021"><img file="EP0833501A2_D0021.tif" /></maths>

A second outlier value detector 210 outputs a second outlier value detecting signal by determining that an impulse component is included in the input sample Y[i][j], when an absolute difference between the input sample Y[i][j] and the mean sample value (A_M) of the small window (W_M) is greater than the mean (kD_M) of an absolute deviation multiplied by the parameter (k), i.e., Y[i][j]-A_M | &gt; kD_M.

A selection control signal generator 212 generates a selection control signal for selecting the output of a trimmed mean filter 214 when the first and second outlier value detecting signals generated by the first and second outlier value detectors 206 and 210 are all generated, and, otherwise, bypassing the input sample without change.

Meanwhile, the trimmed mean filter 214 trims the sample of the small window W_M generated by the window generator 202 using following equation 12, and outputs an output signal y_tr[i][j] by taking the mean of residual samples.<maths id="math0022"><img file="EP0833501A2_D0022.tif" /></maths> wherein,<maths id="math0023"><img file="EP0833501A2_D0023.tif" /></maths> and<maths id="math0024"><img file="EP0833501A2_D0024.tif" /></maths> |T_M| denotes the number of untrimmed samples or the number of samples which are not determined as outlier values. An embodiment of the present invention uses the small window as a trimming window, but the size of the trimming window is variable.

The selector 216 selects the output signal y_tr[i][j] of the trimmed mean filter 214 according to the selection control signal generated by the selection control signal generator 212, i.e., only when the outlier values are detected in both of the first and second outlier value detectors 206 and 210. Otherwise, the selector 216 outputs an output signal (Y=Y_N[i][j]) by bypassing the input sample Y[i][j] output by the window generator 202 without change. The output signal (Y=Y_N[i][j]) is fed back to the window generator 202, and simultaneously output to the contrast enhancer 300 and the colour compensator 500.

Figure 4 is a detailed block diagram of the contrast enhancer 300 shown in Figure 1. Referring to Figure 4, a frame histogram calculator 302 receives a luminance signal (Y) output by the first noise reducer 200 shown in Figure 1 and calculates a histogram in units of one picture. That is, the distribution of the gray level of a frame image is calculated. The picture unit can be a field, but here is set to be a frame. At this time, an input image signal {Y} is comprised of L discrete levels represented by {X₀),X₁,...,X_L-1}.

A frame mean calculator 304 calculates the mean level (X_m) of the luminance signal output by the first noise reducer 200 in units of frame. At this time, X_m ∈ {X₀,X₁,...,X_L-1}. A divider 306 divides the gray level distribution calculated by the frame histogram calculator 302 into the predetermined number (here, two) of subimages {X}_L and {X}_U on the basis of the mean level (X_m) calculated by the frame mean calculator 304, and outputs probability density functions p_L(X_k) and p_U(X_k) of two subimages. The probability density functions p_L(X_k) and p_U(X_k) can be calculated by following equations 13 and 14.<maths id="math0025" num="(13) "><math display="block"><mrow><msub><mrow><mtext mathvariant="italic">p</mtext></mrow><mrow><mtext mathvariant="italic">L</mtext></mrow></msub><mtext>(</mtext><msub><mrow><mtext mathvariant="italic">X</mtext></mrow><mrow><mtext mathvariant="italic">k</mtext></mrow></msub><mtext>) = </mtext><mfrac><mrow><msubsup><mrow><mtext mathvariant="italic">n</mtext></mrow><mrow><mtext mathvariant="italic">k</mtext></mrow><mrow><mtext mathvariant="italic">L</mtext></mrow></msubsup></mrow><mrow><msub><mrow><mtext mathvariant="italic">n</mtext></mrow><mrow><mtext mathvariant="italic">L</mtext></mrow></msub></mrow></mfrac><mtext> ,    </mtext><mtext mathvariant="italic">for k</mtext><mtext> = 0, 1,...,</mtext><mtext mathvariant="italic">m</mtext></mrow></math><img file="EP0833501A2_D0025.tif" /></maths><maths id="math0026" num="(14)"><math display="block"><mrow><msub><mrow><mtext mathvariant="italic">p</mtext></mrow><mrow><mtext mathvariant="italic">U</mtext></mrow></msub><mtext>(</mtext><msub><mrow><mtext mathvariant="italic">X</mtext></mrow><mrow><mtext mathvariant="italic">k</mtext></mrow></msub><mtext>) = </mtext><mfrac><mrow><msubsup><mrow><mtext mathvariant="italic">n</mtext></mrow><mrow><mtext mathvariant="italic">k</mtext></mrow><mrow><mtext mathvariant="italic">u</mtext></mrow></msubsup></mrow><mrow><msub><mrow><mtext mathvariant="italic">n</mtext></mrow><mrow><mtext mathvariant="italic">u</mtext></mrow></msub></mrow></mfrac><mtext> ,    </mtext><mtext mathvariant="italic">for k = m</mtext><mtext>+1, </mtext><mtext mathvariant="italic">m</mtext><mtext>+2,..., </mtext><mtext mathvariant="italic">L</mtext><mtext>-1</mtext></mrow></math><img file="EP0833501A2_D0026.tif" /></maths> wherein L is the number of levels, p_L(X_k) is a probability of a k-th gray level (X_k) in the subimage {X}_L, p_U(X_k) is a probability of a k-th gray level (X_k) in the subimage {X}_U, <i>n</i><maths id="math0027"><math display="inline"><mrow><mfrac linethickness="0"><mrow><mtext mathvariant="italic">L</mtext></mrow><mrow><mtext mathvariant="italic">k</mtext></mrow></mfrac></mrow></math><img file="EP0833501A2_D0027.tif" /></maths><i>and</i> n<maths id="math0028"><math display="inline"><mrow><mfrac linethickness="0"><mrow><mtext mathvariant="italic">U</mtext></mrow><mrow><mtext mathvariant="italic">k</mtext></mrow></mfrac></mrow></math><img file="EP0833501A2_D0028.tif" /></maths> denote the numbers of times in which the level X_k appears in each subimage, and n_L and n_U are the numbers of the respective entire samples in the subimages {X}_L and {X}_U.

A first CDF calculator 308 receives the probability density functions p_L(X_k) of a subimage (hereinafter, called as a first subimage) having samples all being lower than or equal to the mean level (X_m) from the divider 306 and calculates a cumulative density function (CDF) c_L(X_k) using following equation 15.<maths id="math0029"><img file="EP0833501A2_D0029.tif" /></maths>

A second CDF calculator 310 receives the probability density functions p_U(X_k) of a subimage (hereinafter, called as a second subimage) having samples all being greater than the mean level (X_m) from the divider 306 and calculates a cumulative density function c_U(X_k) using following equation 16.<maths id="math0030"><img file="EP0833501A2_D0030.tif" /></maths>

A CDF memory 312 renews the cumulative density functions c_L(X_k) and c_U(X_k) calculated by the first and second CDF calculators 308 and 310 in frame units according to a synchronous signal (SYNC), and provides the previously stored prior-to-one-frame cumulative density functions c_L(X_k) and c_U(X_k) to first and second mappers 316 and 318 during renewal, respectively. Here, the synchronous signal is a field synchronous signal when the picture unit is a field, and is a frame synchronous signal when it is a frame. The CDF memory 312 is used as a buffer.

Meanwhile, a brightness compensator 314 receives the mean level (X_m) output by the frame mean calculator 304, adds a corrected value (Δ) depending on the mean brightness of a signal input to the contrast enhancer 300 to the mean level (X_m) as shown in following equation 17, and outputs a compensated mean level (B_m).<maths id="math0031" num="(17)"><math display="block"><mrow><msub><mrow><mtext>B</mtext></mrow><mrow><mtext>m</mtext></mrow></msub><msub><mrow><mtext> = X</mtext></mrow><mrow><mtext>m</mtext></mrow></msub><mtext> + Δ</mtext></mrow></math><img file="EP0833501A2_D0031.tif" /></maths>

That is, when B_m is a compensated mean level and Δ is a corrected value obtained by a predetermined correction function depending on a mean brightness, the compensated mean level (B_m) becomes a result obtained by adding the corrected value (Δ) to the mean level (X_m). At this time, B_m ⊂ {X₀,X₁,...,X_L-1}.

The corrected value (Δ) is determined by correction functions shown in Figures 5a and 5b. The present invention is not limited to such examples as the correction functions shown in Figures 5a and 5b, and other application examples can exist.

The brightness of enhanced signal (Y_o) is controlled by the corrected value depending on the correction functions shown in Figures 5a and 5b. That is, when the mean level (X_m) is very small, i.e., when an image is very dark, a corrected value (Δ) being larger than 0 is added to the mean level (X_m), and the mean-separate histogram equalization proposed by the present invention is then performed, whereby the mean brightness of the enhanced signal (Y_o) becomes high.

Also, when the mean level (X_m) is very large, i.e., when an image is very bright, a corrected value (Δ) being smaller than 0 is added to the mean level (X_m), and the mean-separate histogram equalization proposed by the present invention is then performed, whereby the mean brightness of the enhanced signal (Y_o) becomes low. Accordingly, when the mean-separate histogram equalization is performed using the mean level (B_m) compensated by a predetermined appropriate corrected value (Δ) according to the mean level (X_m), the image quality of an input image can be remarkably enhanced.

Figures 6a and 6b show the relationship between the mean level (X_m) and the compensated mean level (B_m) to which a corrected value (A) depending on the brightness correction function shown in Figures 5a and 5b is added.

Meanwhile, the first mapper 316 shown in Figure 4 receives the cumulative density function c_L(X_k) calculated by the first CDF calculator 308, the signal (X_k) output from the first noise reducer 200, and the compensated mean level (B_m) output by the brightness compensator 314, and maps the samples {X}_L of the first subimage to a gray level ranging from 0 to B_m according to the cumulative density function.

The second mapper 318 receives the cumulative density function c_U(X_k) calculated by the second CDF calculator 310, the signal (X_k) output by the first noise reducer 200, and the compensated mean level (B_m) output by the brightness compensator 314, and maps the samples {X}_U of the second subimage to a gray level ranging from B_m' to X_L-1 according to the cumulative density function.

The outputs mapped by the first and second mappers 316 and 318 are expressed by following equation 18, and B_m' is expressed by following equation 19.<maths id="math0032"><img file="EP0833501A2_D0032.tif" /></maths><maths id="math0033" num="(19)"><math display="block"><mrow><msub><mrow><mtext mathvariant="italic">B</mtext></mrow><mrow><mtext mathvariant="italic">m</mtext></mrow></msub><msup><mrow><mtext mathvariant="italic"> </mtext></mrow><mrow><mtext>'</mtext></mrow></msup><mtext> = </mtext><msub><mrow><mtext mathvariant="italic">B</mtext></mrow><mrow><mtext mathvariant="italic">m</mtext></mrow></msub><mtext> + </mtext><msub><mrow><mtext mathvariant="italic">X</mtext></mrow><mrow><mtext mathvariant="italic">L</mtext></mrow></msub><msub><mrow><mtext>​</mtext></mrow><mrow><mtext>-1</mtext></mrow></msub><mtext>/(</mtext><mtext mathvariant="italic">L</mtext><mtext>-1)</mtext></mrow></math><img file="EP0833501A2_D0033.tif" /></maths> wherein B_m' denotes a first gray level which is mapped in a higher region than the compensated mean level (B_m).

Accordingly, the equation 18 shows the results of mapping input samples to (0,B_m) when they are equal to or smaller than the mean level (X_m) and mapping the input samples to (B_m',X_L-1) when they are larger than the mean level (X_m).

Also, when a corrected value is larger than 0 (i.e., Δ &gt; 0), the equalized output (Y_o) becomes bright, and when it is smaller than 0 (i.e., Δ &lt; 0), the equalized output (Y_o) becomes dark. As the Δ increases, a dynamic range at a lower region will be enhanced, and as the Δ decreases, a dynamic range at an upper region will be enhanced. A predetermined appropriate compensated mean level (B_m) depending on the mean level (X_m) greatly enhances the image quality of an input image.

A comparator 320 compares the signal (X_k) output by the first noise reducer 200 with the mean level (X_m) output by the frame mean calculator 304 and generates a selection control signal. A first selector 322 selects the first mapper 316 when the signal (X_k) output by the first noise reducer 200 is lower than the mean level (X_m), and otherwise, selects the second mapper 318.

Here, the signal (X_k) input to the first and second mappers 316 and 318 is a signal of the frame next to a frame corresponding to the cumulative density function value output by the CDF memory 312. Thus, a frame memory for delaying a signal output by the first noise reducer 200 by one frame can be additionally provided to input a signal of the same frame as that of the cumulative density function output by the CDF memory 312 to the first and second mappers 316 and 318. However, the frame memory can be omitted using the characteristic that there is a high correlation between adjacent frames, so that hardware is reduced.

Furthermore, without separately using the frame histogram calculator 302 and the CDF calculators 308 and 310, the gray level distribution of an image signal of each subgroup by the CDF calculators 308 and 310 without the frame histogram calculator 302 is calculated, and a CDF can be calculated on the basis of the result.

Meanwhile, a subtracter 324 and an adder 334 perform a function of controlling the gain of the enhanced signal (Y_o). Here, a basic concept of the gain control is that a variation of the maximum gray level of an input signal (Y) is restricted according to the degree of contrast enhancement when a contrast is enhanced using the mean-separate histogram equalization.

First, the relationship between the input signal (Y=X_k) and the enhanced signal (Y_o) will be expressed by following equation 20 or 21;<maths id="math0034" num="(20)"><math display="block"><mrow><msub><mrow><mtext>Y</mtext></mrow><mrow><mtext>o</mtext></mrow></msub><msub><mrow><mtext> = X</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><msub><mrow><mtext> + Δ</mtext></mrow><mrow><mtext>k</mtext></mrow></msub></mrow></math><img file="EP0833501A2_D0034.tif" /></maths> or<maths id="math0035" num="(21)"><math display="block"><mrow><msub><mrow><mtext>Y</mtext></mrow><mrow><mtext>o</mtext></mrow></msub><msub><mrow><mtext> - X</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><msub><mrow><mtext> = Δ</mtext></mrow><mrow><mtext>k</mtext></mrow></msub></mrow></math><img file="EP0833501A2_D0035.tif" /></maths> wherein Δ_k is the amount of variation (the degree of enhancement) made by the mean-separate histogram equalization when an input sample is X_k, i.e., a difference between the level of the input signal (Y=X_k) and the level (Y_o) mapped to a new gray level by the mean-separate histogram equalization.

In order to prevent excess enhancement due to histogram equalization, the amount of variation Δ_k according to the present invention is restricted as follows.<maths id="math0036" num="(22)"><math display="block"><mrow><mfenced open="|" close="|"><mrow><msub><mrow><mtext>Δ</mtext></mrow><mrow><mtext>k</mtext></mrow></msub></mrow></mfenced><msub><mrow><mtext> ≤ γ · g(X</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><mtext>)</mtext></mrow></math><img file="EP0833501A2_D0036.tif" /></maths> wherein g(X_k) denotes a maximum bounding function, the g(X_k) is a function of the input signal (Y=X_k) and always has a positive value (i.e., g(X_k)≥0), and γ(γ≥0) is a gain control parameter. The above equation 22 can be expressed by following equation 23 as the same equation.<maths id="math0037" num="(23)"><math display="block"><mrow><msub><mrow><mtext>-r g(X</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><msub><mrow><mtext>) ≤ Δ</mtext></mrow><mrow><mtext>i</mtext></mrow></msub><msub><mrow><mtext> ≤ r·g(X</mtext></mrow><mrow><mtext>k</mtext></mrow></msub><mtext>)</mtext></mrow></math><img file="EP0833501A2_D0037.tif" /></maths>

The concept of restricting the amount of variation Δ_k expressed as the equation 22 is related to the Weber's ratio. In fact, if g(X_k) is equal to X_k, the following equation 24 can be obtained from the above equation 22.<maths id="math0038" num="(24)"><math display="block"><mrow><mfrac><mrow><mfenced open="|" close="|"><mrow><msub><mrow><mtext>Δ</mtext></mrow><mrow><mtext mathvariant="italic">k</mtext></mrow></msub></mrow></mfenced></mrow><mrow><msub><mrow><mtext mathvariant="italic">X</mtext></mrow><mrow><mtext mathvariant="italic">k</mtext></mrow></msub></mrow></mfrac><mtext>≤γ</mtext></mrow></math><img file="EP0833501A2_D0038.tif" /></maths> wherein<maths id="math0039"><math display="block"><mrow><mfrac><mrow><mfenced open="|" close="|"><mrow><msub><mrow><mtext>Δ</mtext></mrow><mrow><mtext mathvariant="italic">k</mtext></mrow></msub></mrow></mfenced></mrow><mrow><msub><mrow><mtext mathvariant="italic">X</mtext></mrow><mrow><mtext mathvariant="italic">k</mtext></mrow></msub></mrow></mfrac></mrow></math><img file="EP0833501A2_D0039.tif" /></maths> is the amount corresponding to the Weber's ratio. The Weber's ratio is an experimental fact that human beings feel that, when X₁ is changed by γX₁ and X₂ is changed by γX₂, the degrees of the variations are the same. Thus, the concept of the gain control applied to the present invention is to control the gain, i.e., the enhancement degree of an enhanced signal using the mean-separate histogram having a brightness compensating function on the basis of the Weber's ratio.

The subtracter 324 shown in Figure 4 subtracts the input signal (Y=X_k) output by the first noise reducer 200 from the enhanced signal (Y_o) output by the first selector 322 and obtains the amount of variation (Δ_k) made by the mean-separate histogram equalization having the brightness compensation function.

A gain feature determiner 326 restricts enhancement of the input signal (X_k) by outputting a maximum bounding function g(X_k) being a function of the input signal (X_k). For instance, when the g(X_k) is K1 and the K1 is a constant, the g(X_k) restricts input enhancement by identical amount regardless of an input gray level value. The maximum bounding function g(X_k) can be equal to aX_k or <maths id="math0040"><math display="inline"><mrow><mtext mathvariant="italic">a</mtext><mtext></mtext><msqrt><msub><mrow><mtext mathvariant="italic">X</mtext></mrow><mrow><mtext mathvariant="italic">k</mtext></mrow></msub></msqrt></mrow></math><img file="EP0833501A2_D0040.tif" /></maths> (here, a is a constant), and differently restricts the enhanced amount of an input image according to the input gray level value.

A second selector 328 compares the input signal (X_k) with the mean level (X_m) output by the frame mean calculator 304, and selects a first gain control parameter (γ_L) when the input signal (X_k) is lower than or equal to the mean level (X_m) and otherwise, selects a second gain control parameter (γ_U). Here, the first gain control parameter (γ_L) is a parameter for a first subimage signal, and the second gain control parameter (γ_U) is a parameter for a second subimage signal. Here, the first and second gain control parameters (γ_L) and (γ_U) can be given as the same value to the input signal in contrast that different gain control parameters are provided to each subimage.

A multiplier 330 multiplies the value of the maximum bounding function g(X_k) output by the gain feature determiner 326 by a gain control parameter (γ) selected by the second selector 328 and outputs a limit value (γ · g(X_k)). Here, the limit value γ · g(X_k) is a characteristic of the gain control according to the present invention.

A limiter 332 compares the variation amount (Δ_k) output by the subtracter 324 with the limit value γ · g(X_k) output by the multiplier 330, restricts the variation amount (Δ_k), and outputs a restricted variation amount (Δ_k') like the following equation 25.<maths id="math0041"><img file="EP0833501A2_D0041.tif" /></maths>

That is, when an absolute value of the variation amount (Δ_k) is equal to or smaller than γ · g(X_k), the variation amount (Δ_k) is used as the restricted variation amount (Δ_k'). When the variation amount (Δ_k) is greater than γ · g(X_k), the variation amount (Δ_k) is restricted to the γ · g(X_k). When the variation amount (Δ_k) is smaller than -γ · g(X_k), the variation amount (Δ_k) is restricted to the -γ · g(X_k), thereby controlling the gain of the enhanced signal (Y_o).

The adder 334 adds the input signal (X_k) to the restricted variation amount (Δ_k') output by the limiter 332 and outputs an output signal (Y_H) as shown in the following equation 26. Thus, the final output signal (Y_H) of the contrast enhancer 300 can be expressed as follows.<maths id="math0042" num="(26)"><math display="block"><mrow><msub><mrow><mtext mathvariant="italic">Y</mtext></mrow><mrow><mtext mathvariant="italic">H</mtext></mrow></msub><mtext> = </mtext><msub><mrow><mtext mathvariant="italic">X</mtext></mrow><mrow><mtext mathvariant="italic">k</mtext></mrow></msub><msub><mrow><mtext> + Δ</mtext></mrow><mrow><mtext mathvariant="italic">k</mtext></mrow></msub><msup><mrow><mtext> </mtext></mrow><mrow><mtext>'</mtext></mrow></msup></mrow></math><img file="EP0833501A2_D0042.tif" /></maths>

Figure 7 is a detailed block diagram of the local contrast enhancer 400 shown in Figure 1, wherein an M×N window generator 402 receiving the output (Y_H) of the contrast enhancer 300 shown in Figure 1 generates an M×N window (W) which can be expressed by the following equation 27, and outputs it to an M×N low-pass filter (LPF) 404.<maths id="math0043"><img file="EP0833501A2_D0043.tif" /></maths>

Also, a middle sample (x) of a center line of the M×N window (W) is input to a subtracter 406 and an adder 416. Here, the middle sample (x) of the center line is an input sample for improving a contrast.

Figure 8 is a detailed circuit diagram of the M×N window generator 402 when both M and N are equal to 3. Referring to Figure 8, D and H denote a sample delay and a line memory, respectively. The output signal (m) of the M×N LPF 404 shown in Figure 7 can be represented by the following equation 28.<maths id="math0044"><img file="EP0833501A2_D0044.tif" /></maths> wherein b_ij is a predetermined coefficient, and<maths id="math0045"><img file="EP0833501A2_D0045.tif" /></maths> corresponds to an impulse response of the M×NLPF 404.

A subtracter 406 subtracts the output signal (m) of the M×N LPF 404 from the middle sample value (x). An absolute value calculator 408 calculates an absolute value of the output of the subtracter 406, and a weighting unit 410 outputs a value obtained by weighting the output of the absolute value calculator 408 using a predetermined weighting function f(|x-m|). A first multiplier 412 multiplies a predetermined parameter (β) by the output of the weighting unit 410.

A second multiplier 414 multiplies the output of the subtracter 406 by the output of the first multiplier 412 and outputs the multiplied result β·f(|x-m|) (x-m). An adder 416 adds the multiplied result to the middle sample (x) of the M×N window generator 402.

At this time, a difference between the middle sample (x) of the M×Nwindow generator 402 and the output signal (m) of the M×N LPF 404 can be defined as a contrast visually felt by a human being, i.e., a local contrast. In other words, a place where the value |x-m| is large can be called a region having a high contrast, and a place where the value |x-m| is small can be called a region having a low contrast. The local contrast is enhanced by amplifying the value |x-m| by f(|x-m|) according to the above-defined local contrast and adding the result to the original signal (x). The output (Y') of the adder 416 will be expressed by following equation 29.<maths id="math0046" num="(29)"><math display="block"><mrow><mtext>Y' = x + β·f(|x-m|) (x-m)</mtext></mrow></math><img file="EP0833501A2_D0046.tif" /></maths> wherein the function f() being a weighting function is a function of f|x-m|. Various types of enhancement characteristics for the local contrast can be realized by appropriately selecting a weighting function. Also, β is a parameter for adjusting the amount of enhancement of an entire local contrast.

For example, when f(|x-m|) equals 0, Y' equals x, i.e., there is no contrast enhancement effect. Thus, given that when |x-m| is smaller than T, f(|x-m|) is K2, and that when |x-m| is greater than T, f(|x-m|) is 0,(here, T and K2 are constants), the local contrast is amplified by K2 times in a region where the local contrast is small (i.e., |x-m| &lt; T). On the other hand, in a region where it is already determined that the local contrast is large, adaptive local contrast enhancement of bypassing an input sample is performed.

Figure 9 show examples of a weighting function for determining a weighted value of a local contrast according to |x-m|. Various contrast enhancement characteristics can be obtained by using the weighting functions indicated by (a) and (b).

Meanwhile, a colour compensation method will be described with reference to Figures 10 and 11 before describing the colour compensator 500 shown in Figure 1.

A colour signal C is given as R, G and B signals, the luminance signal Y shown in the equation 1 is given as a1R+a2G+a3B, and Y is converted into Y' through the contrast enhancer 300 and the local contrast enhancer 400. At this time, the object of the colour compensation is to properly change the original colour signal C=(R,G,B) according to the change of Y to Y'.

The basic concept of the colour compensation according to the present invention is to change a given colour to the colour direction in a (R,G,B) space.

First, where Y equals a1R+a2G+a3B, R, G and B each having a constant value Y form a plane in the (R,G,B) space. That is, every colour signal put on the plane where Y equals a1R+a2G+a3B has an identical luminance value. The change of luminance from Y to Y' involves the fact that the given colour C is moved to a position on the Y' plane as shown in Figure 10. At this time, in the present invention, suppose that the colours C and C' have the same colour direction. Here, when C' equals (R',G',B'), it means that a straight line OC coincides with a straight line OC'. Thus, the compensated colour C' on the Y' plane is obtained by detecting an intersecting point of the straight line OC and the Y' plane. Briefly, the original colour signal C shown in Figure 10 is mapped into the new colour signal C' when the luminance value is changed from Y to Y', which is the intersecting point of the straight line OC and the Y' plane.

Now, (l,m,n) is defined as a directional cosine of the given colour C in order to obtain the colour C', which will be expressed by following equation 30;<maths id="math0047" num="(30)"><math display="block"><mrow><mtext>l = R/r, m = G/r, n = B/r,</mtext></mrow></math><img file="EP0833501A2_D0047.tif" /></maths> wherein r equals <maths id="math0048"><math display="inline"><mrow><msqrt><msup><mrow><mtext mathvariant="italic">R</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><mtext>+</mtext><msup><mrow><mtext mathvariant="italic">G</mtext></mrow><mrow><mtext>2</mtext></mrow></msup><mtext>+</mtext><msup><mrow><mtext mathvariant="italic">B</mtext></mrow><mrow><mtext>2</mtext></mrow></msup></msqrt></mrow></math><img file="EP0833501A2_D0048.tif" /></maths> . Similarly, the directional cosines l', m' and n' of the output colour signals R', G' and B' can be expressed by following equation 31;<maths id="math0049" num="(31)"><math display="block"><mrow><mtext>l' = R'/r', m' = G'/r', n' = B'/r',</mtext></mrow></math><img file="EP0833501A2_D0049.tif" /></maths> wherein r' equals <maths id="math0050"><math display="inline"><mrow><msqrt><mtext mathvariant="italic">R</mtext><msup><mrow><mtext>' </mtext></mrow><mrow><mtext>2</mtext></mrow></msup><mtext>+</mtext><mtext mathvariant="italic">G</mtext><msup><mrow><mtext>' </mtext></mrow><mrow><mtext>2</mtext></mrow></msup><mtext>+</mtext><mtext mathvariant="italic">B</mtext><msup><mrow><mtext>' </mtext></mrow><mrow><mtext>2</mtext></mrow></msup></msqrt></mrow></math><img file="EP0833501A2_D0050.tif" /></maths> . To make the two types of colours have the same colour direction, the relationships shown in following equations 32 and 33 should be accomplished.<maths id="math0051" num="(32)"><math display="block"><mrow><mtext>l = l', m = m', n =n'</mtext></mrow></math><img file="EP0833501A2_D0051.tif" /></maths><maths id="math0052" num="(33)"><math display="block"><mrow><mtext>R'/r' = R/r, G'/r' = G/r, B'/r' = B/r</mtext></mrow></math><img file="EP0833501A2_D0052.tif" /></maths> Accordingly, the following equations 34, 35 and 36 are obtained.<maths id="math0053" num="(34)"><math display="block"><mrow><mtext mathvariant="italic">R'</mtext><mtext>= </mtext><mfrac><mrow><mtext mathvariant="italic">r'</mtext></mrow><mrow><mtext mathvariant="italic">r</mtext></mrow></mfrac><mtext mathvariant="italic">R</mtext></mrow></math><img file="EP0833501A2_D0053.tif" /></maths><maths id="math0054" num="(35)"><math display="block"><mrow><mtext mathvariant="italic">G'</mtext><mtext>= </mtext><mfrac><mrow><mtext mathvariant="italic">r'</mtext></mrow><mrow><mtext mathvariant="italic">r</mtext></mrow></mfrac><mtext mathvariant="italic">G</mtext></mrow></math><img file="EP0833501A2_D0054.tif" /></maths><maths id="math0055" num="(36)"><math display="block"><mrow><mtext mathvariant="italic">B'</mtext><mtext>= </mtext><mfrac><mrow><mtext mathvariant="italic">r'</mtext></mrow><mrow><mtext mathvariant="italic">r</mtext></mrow></mfrac><mtext mathvariant="italic">B</mtext></mrow></math><img file="EP0833501A2_D0055.tif" /></maths> When the relationships given by equations 34, 35 and 36 are substituted for Y'=a1R'+awG'+a3B', following equations 37, 38 and 39 are obtained;<maths id="math0056" num="(37)"><math display="block"><mrow><mtext mathvariant="italic">Y'</mtext><mtext> = </mtext><msub><mrow><mtext mathvariant="italic">a</mtext></mrow><mrow><mtext>1</mtext></mrow></msub><mfrac><mrow><mtext mathvariant="italic">r'</mtext></mrow><mrow><mtext mathvariant="italic">r</mtext></mrow></mfrac><mtext></mtext><mtext mathvariant="italic">R</mtext><mtext> + </mtext><msub><mrow><mtext mathvariant="italic">a</mtext></mrow><mrow><mtext>2</mtext></mrow></msub><mfrac><mrow><mtext mathvariant="italic">r'</mtext></mrow><mrow><mtext mathvariant="italic">r</mtext></mrow></mfrac><mtext></mtext><mtext mathvariant="italic">G</mtext><mtext> + </mtext><msub><mrow><mtext mathvariant="italic">a</mtext></mrow><mrow><mtext>3</mtext></mrow></msub><mtext></mtext><mfrac><mrow><mtext mathvariant="italic">r'</mtext></mrow><mrow><mtext mathvariant="italic">r</mtext></mrow></mfrac><mtext></mtext><mtext mathvariant="italic">B</mtext></mrow></math><img file="EP0833501A2_D0056.tif" /></maths><maths id="math0057" num="(38)"><math display="block"><mrow><mtext>Y' = </mtext><mfrac><mrow><mtext mathvariant="italic">r'</mtext></mrow><mrow><mtext mathvariant="italic">r</mtext></mrow></mfrac><mtext> (</mtext><msub><mrow><mtext mathvariant="italic">a</mtext></mrow><mrow><mtext>1</mtext></mrow></msub><mtext mathvariant="italic">R</mtext><mtext> + </mtext><msub><mrow><mtext mathvariant="italic">a</mtext></mrow><mrow><mtext>2</mtext></mrow></msub><mtext mathvariant="italic">G</mtext><mtext> + </mtext><msub><mrow><mtext mathvariant="italic">a</mtext></mrow><mrow><mtext>3</mtext></mrow></msub><mtext mathvariant="italic">B</mtext><mtext>)</mtext></mrow></math><img file="EP0833501A2_D0057.tif" /></maths><maths id="math0058" num="(39)."><math display="block"><mrow><mtext>Y' = </mtext><mfrac><mrow><mtext mathvariant="italic">r'</mtext></mrow><mrow><mtext mathvariant="italic">r</mtext></mrow></mfrac><mtext mathvariant="italic">Y</mtext></mrow></math><img file="EP0833501A2_D0058.tif" /></maths>

Accordingly, such a result as the following equation 40 is obtained.<maths id="math0059" num="(40)"><math display="block"><mrow><mfrac><mrow><mtext mathvariant="italic">Y'</mtext></mrow><mrow><mtext mathvariant="italic">Y</mtext></mrow></mfrac><mtext> = </mtext><mfrac><mrow><mtext mathvariant="italic">r'</mtext></mrow><mrow><mtext mathvariant="italic">r</mtext></mrow></mfrac></mrow></math><img file="EP0833501A2_D0059.tif" /></maths>

Meanwhile, Equations 34 through 36 can be also expressed as following Equations 41 through 53 by using the result of the Equation 40.<maths id="math0060" num="(41)"><math display="block"><mrow><mtext mathvariant="italic">R'</mtext><mtext> = </mtext><mfrac><mrow><mtext mathvariant="italic">Y'</mtext></mrow><mrow><mtext mathvariant="italic">Y</mtext></mrow></mfrac><mtext mathvariant="italic">R</mtext></mrow></math><img file="EP0833501A2_D0060.tif" /></maths><maths id="math0061" num="(42)"><math display="block"><mrow><mtext mathvariant="italic">G'</mtext><mtext> = </mtext><mfrac><mrow><mtext mathvariant="italic">Y'</mtext></mrow><mrow><mtext mathvariant="italic">Y</mtext></mrow></mfrac><mtext mathvariant="italic">G</mtext></mrow></math><img file="EP0833501A2_D0061.tif" /></maths><maths id="math0062" num="(43)"><math display="block"><mrow><mtext mathvariant="italic">B'</mtext><mtext> = </mtext><mfrac><mrow><mtext mathvariant="italic">Y'</mtext></mrow><mrow><mtext mathvariant="italic">Y</mtext></mrow></mfrac><mtext mathvariant="italic">B</mtext></mrow></math><img file="EP0833501A2_D0062.tif" /></maths> Consequently, C' can be obtained as follows.<maths id="math0063" num="(44)"><math display="block"><mrow><mtext>C' = (R' ,G' ,B')</mtext></mrow></math><img file="EP0833501A2_D0063.tif" /></maths><maths id="math0064" num="(45)"><math display="block"><mrow><mtext>C' = (qR,qG,qB)</mtext></mrow></math><img file="EP0833501A2_D0064.tif" /></maths> wherein q equals Y'/Y which is a ratio between the original luminance signal and the resultant luminance signal. Since this means that the ratio of a luminance change is equal to that of a colour change, the present invention performs colour compensation by changing a colour value according to the luminance change.

The colour compensation can be easily performed on other colour systems using the result of Equation 45. That is, for instance, the (Y,U,V) signals given in Equation 1, being the results of the colour compensation given in Equation 45, should be converted into (qY,qU,qV) as shown in following Equations 46 to 51:<maths id="math0065" num="(46)"><math display="block"><mrow><msub><mrow><mtext>U' = b</mtext></mrow><mrow><mtext>1</mtext></mrow></msub><msub><mrow><mtext>R' + b</mtext></mrow><mrow><mtext>2</mtext></mrow></msub><msub><mrow><mtext>G' + b</mtext></mrow><mrow><mtext>3</mtext></mrow></msub><mtext>B'</mtext></mrow></math><img file="EP0833501A2_D0065.tif" /></maths><maths id="math0066" num="(47)"><math display="block"><mrow><msub><mrow><mtext>U = q(b</mtext></mrow><mrow><mtext>1</mtext></mrow></msub><msub><mrow><mtext>R + b</mtext></mrow><mrow><mtext>2</mtext></mrow></msub><msub><mrow><mtext>G + b</mtext></mrow><mrow><mtext>3</mtext></mrow></msub><mtext>B)</mtext></mrow></math><img file="EP0833501A2_D0066.tif" /></maths><maths id="math0067" num="(48)"><math display="block"><mrow><mtext>U = qU</mtext></mrow></math><img file="EP0833501A2_D0067.tif" /></maths> and,<maths id="math0068" num="(49)"><math display="block"><mrow><msub><mrow><mtext>V' = c</mtext></mrow><mrow><mtext>1</mtext></mrow></msub><msub><mrow><mtext>R' + c</mtext></mrow><mrow><mtext>2</mtext></mrow></msub><msub><mrow><mtext>G' + c</mtext></mrow><mrow><mtext>3</mtext></mrow></msub><mtext>B'</mtext></mrow></math><img file="EP0833501A2_D0068.tif" /></maths><maths id="math0069" num="(50)"><math display="block"><mrow><msub><mrow><mtext>V = q(c</mtext></mrow><mrow><mtext>1</mtext></mrow></msub><msub><mrow><mtext>R + c</mtext></mrow><mrow><mtext>2</mtext></mrow></msub><msub><mrow><mtext>G + c</mtext></mrow><mrow><mtext>3</mtext></mrow></msub><mtext>B)</mtext></mrow></math><img file="EP0833501A2_D0069.tif" /></maths><maths id="math0070" num="(51)."><math display="block"><mrow><mtext>V = qV</mtext></mrow></math><img file="EP0833501A2_D0070.tif" /></maths>

Now, colour compensation will be performed to prevent colour saturation due to the above-described image quality enhancement method.

The luminance ratio is given as q, and when the colour compensation is made by a compensating line (R',G',B')=q(R,G,B) shown in Figure 11, a colour signal between a b(=Max/q) level and a maximum (Max) level is compensated to a maximum value (Max), which involves being saturated. That is because the colours between b and Max are consequently mapped to the Max and are not distinguished from each other, referring to the above-described compensating method.

In order to prevent the colour saturation, the compensating line (R',G',B')=q(R,G,B) is approximated to a compensating line (R',G',B')=A(R,G,B)+K, for a colour signal between a(=αb) level and a maximum level (Max). Here, a parameter (α) is greater than or equal to 0 and lower than or equal to 1, and A and K can be expressed by following Equations 52 and 53;<maths id="math0071" num="(52)"><math display="block"><mrow><mtext mathvariant="italic">A</mtext><mtext> = </mtext><mfrac><mrow><mtext mathvariant="italic">q</mtext><mfenced open="(" close=")"><mrow><mtext>1-α</mtext></mrow></mfenced></mrow><mrow><mfenced open="(" close=")"><mrow><mtext mathvariant="italic">q</mtext><mtext>-1</mtext></mrow></mfenced></mrow></mfrac></mrow></math><img file="EP0833501A2_D0071.tif" /></maths> and,<maths id="math0072" num="(53)."><math display="block"><mrow><mtext mathvariant="italic">K</mtext><mtext> = </mtext><mfrac><mrow><mfenced open="(" close=")"><mrow><mtext mathvariant="italic">q</mtext><mtext>-1</mtext></mrow></mfenced><mtext>α</mtext></mrow><mrow><mtext mathvariant="italic">q</mtext><mtext>-1</mtext></mrow></mfrac><mtext></mtext><mtext mathvariant="italic">Max</mtext></mrow></math><img file="EP0833501A2_D0072.tif" /></maths>

Figure 12 is a detailed circuit diagram of the colour compensator 500 shown in Figure 1 according to an embodiment of the present invention, wherein an operator 502 operates a ratio (q) between a signal Y output by the first noise reducer 200 of Figure 1 and a signal Y' output by the local contrast enhancer 400, i.e., the Y'/Y.

A first multiplier 504 multiplies a signal U output by the second noise reducer 220 of Figure 1 by the ratio (q) output by the operator 502 and outputs a signal U'. A second multiplier 506 multiplies a signal V output by the third noise reducer 240 of Figure 1 by the ratio (q) output by the operator 502 and outputs a signal V'.

Figure 13 is a detailed circuit diagram of the colour compensator 500 shown in Figure 1 according to another embodiment of the present invention, wherein an operator 512 operates a ratio (q) between a signal Y output by the first noise reducer 200 of Figure 1 and a signal Y' output by the local contrast enhancer 400, i.e., the Y'/Y.

First approximate value regulators 514 and 516 are provided to prevent colour compensation using the approximated compensating line shown in Figure 11. The first approximate value regulator 514 receives the signal U output by the second noise reducer 220 of Figure 1, the ratio (q) output by the operator 512, and a parameter (α), and outputs a signal U' compensated by the compensating line (R',G',B')=q(R,G,B) when the input signal U is between the minimum level and the a(=αb) level and outputs a signal U' compensated by the approximated compensating line (R',G',B')=A(R,G,B)+K when the input signal U is between the "a" level and the maximum level (Max).

At this time, the signal U can be compensated either by the compensating line (R',G',B')=q(R,G,B) or by the approximated compensating line (R',G',B')=A(R,G,B)+K, depending on the parameter value (a). That is, when the parameter value (α) is 1, the signal U is compensated by the compensating line (R',G',B')=q(R,G,B). When the parameter value (α) is greater than 0 and smaller than 1, the signal U being between the minimum level and the"a"(=αb) level is compensated by the compensating line (R',G',B')=q(R,G,B), and the signal U being between the "a" level and the maximum level is compensated by the approximated compensating line (R',G',B')=A(R,G,B)+K.

The second approximate value regulator 516 receives the signal V output by the third noise reducer 240 of Figure 1, the ratio (q) output by the operator 512, and a parameter (α), and outputs a signal V' compensated by the compensating line (R',G',B')=q(R,G,B) when the input signal V is between the minimum level and the "a" level and outputs a signal V' compensated by the approximated compensating line (R',G',B')=A(R,G,B)+K when the input signal V is between the "a" level and the maximum level (Max).

Thus, embodiments of the present invention are applicable to a wide-ranging field associated with image quality enhancement for an image signal. That is, the present invention can be applied to broadcasting apparatuses, radar signal processing systems, medical equipment, electric home appliances, etc.

As described above, embodiments of the present invention, which considers a corrected value depending on the mean brightness of an input image and uses the mean-separate histogram equalization having the gain control function for preventing excessive enhancement, effectively reduce an abrupt change in brightness and an artifact generated during conventional histogram equalization, thereby enhancing a contrast and greatly improving the image quality of an input image. Furthermore, embodiments of the present invention effectively remove impulse noise by increasing reliability through a dual impulse detecting method, thus improving the image quality. The image quality can also be enhanced through enhancement of a local contrast. Also, when a luminance is changed by a predetermined process for contrast enhancement, embodiments of the present invention vary a colour value according to the change, thereby providing an undistorted colour signal.

The reader's attention is directed to all papers and documents which are filed concurrently with or previous to this specification in connection with this application and which are open to public inspection with this specification, and the contents of all such papers and documents are incorporated herein by reference.

All of the features disclosed in this specification (including any accompanying claims, abstract and drawings), and/or all of the steps of any method or process so disclosed, may be combined in any combination, except combinations where at least some of such features and/or steps are mutually exclusive.

Each feature disclosed in this specification (including any accompanying claims, abstract and drawings), may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise. Thus, unless expressly stated otherwise, each feature disclosed is one example only of a generic series of equivalent or similar features.

The invention is not restricted to the details of the foregoing embodiment(s). The invention extends to any novel one, or any novel combination, of the features disclosed in this specification (including any accompanying claims, abstract and drawings), or to any novel one, or any novel combination, of the steps of any method or process so disclosed.