Motion detection device and noise reduction device using that

Mobile detection device and noise elimination device using the device

Technical field

The present invention relates to image signal processing technology in a television receiver.

Background technique

As a method of removing noise from an image input signal mixed with noise, there is a frame loop type noise removal method. For example, there is a method of obtaining the difference between the frame-delayed signal of the image output signal (with noise removed) and the image input signal, and multiplying the difference by the desired cyclic coefficient (more than 0 but less than 1), This value is added and subtracted from the original image input signal, thereby removing noise.

In addition, the traditional technology of motion detection and noise cancellation is described in Japanese Patent Application Publication No. 2002-93834 and "Transistor Technology Special Issue" (Matsui Shunya, No. 52, 1999, CQ Publishing Co., Ltd., pages 89-92) It is also described in et al.

Below, an example of the traditional technology.

18A and 18B show an example of a conventional frame loop type noise canceling device. In FIG. 18A, the frame memory 1801 performs frame delay on the image output signal (noise removed) 1807. The subtractor 1802 obtains the frame difference between the output of the frame memory 1801 and the image input signal 1806. This difference is input to the movement detection and loop coefficient generation circuit 1803 that performs movement detection and loop coefficient generation. The motion detection and loop coefficient generation circuit 1803 generates a loop coefficient based on the difference.

Fig. 18B shows an example of the relationship between this difference and the cyclic coefficient.

In FIG. 18B, the horizontal axis 1810 represents the difference value, the vertical axis 1811 represents the cyclic coefficient, and the curve 1812 represents the relationship between the difference value and the cyclic coefficient. As shown by the curve 1812, the movement detection and cyclic coefficient generation circuit 1803 has the characteristic of making the cyclic coefficient non-linearly smaller as the difference increases. This is to use the correlation between the increase in the movement and the increase in the difference to perform movement detection. The cyclic coefficient and the frame difference value thus obtained are multiplied in the multiplier 1804. Furthermore, the adder and subtractor 1805 adds and subtracts the output of the multiplier 1804 from the image input signal 1806 according to the sign of the frame difference, thereby obtaining an image output signal 1807 with noise removed.

In the above-mentioned traditional method, the effect of noise cancellation is increased by increasing the loop coefficient. However, in this case, the disadvantage of the attenuation tail caused by the frame difference will occur for the moving part. Therefore, there is a method of performing nonlinear processing based on the frame difference value, or changing the loop coefficient based on the result of motion detection. However, because traditional motion detection uses frame difference as it is, or uses horizontal and vertical edge detection, it cannot distinguish between noise and movement for moving parts that are at or below the noise level. . To remove noise in this state, the phenomenon of attenuating the tail will occur. Therefore, in such a state, from the viewpoint of preventing the disadvantages of the attenuation tail, the noise cancellation effect is limited.

In addition, there is a method of either performing nonlinear processing based on the frame difference value, or changing the loop coefficient based on the result of motion detection. This method is described in the non-patent literature mentioned above. However, traditional motion detection uses the frame or field difference as it is. In this case, no distinction can be made between noise and movement for moving parts at or below the noise level. In this state, if noise is to be eliminated, afterimages will occur. Therefore, in such a state, the noise canceling effect is limited from the viewpoint of preventing afterimages.

In the above configuration, the difference caused by the movement in the frame difference is the same or lower than the difference caused by the noise. Since the two cannot be distinguished, the occurrence of attenuation tails can be prevented. From the point of view of the purpose, the inability to increase the cycle coefficient will limit the noise elimination effect.

In addition, in order to appropriately improve the noise cancellation effect for dark parts where the amount of noise is likely to increase, a method of controlling the noise cancellation level based on the information from the AGC and the input signal level can also be considered. However, it is necessary to add a new control circuit in order to operate in conjunction with the AGC circuit. In addition, in the method of using the input signal level itself, the control level is easily changed in the spatial direction due to the influence of noise.

Furthermore, in the case of appropriately combining cyclic noise cancellation and acyclic noise cancellation using a spatially oriented filter based on the motion detection result, when the conventional discontinuous motion detection result is used, it is accompanied by discontinuity in the spatial direction. The processing of, it is easy to produce an unnatural state on the surface, and for the animation containing the outline, because the filter in the space direction is added, the outline is likely to be blurred during the animation, so an additional outline trimming circuit is required.

In addition, generally the still image part and the dynamic image part appear in a certain consistency in the image. Therefore, in this field, it is very possible to use the same loop coefficient between adjacent pixels from the result of motion detection, especially when the loop coefficient is increased, which is easy to produce such as the impression that the image appears to be pasted. Wait for the unnatural state on the surface. In addition, the threshold value in the motion detection becomes larger when the electric field is weak, and an increase in the cycle coefficient is likely to produce an unnatural feeling that a thin film is formed on the entire image.

Summary of the invention

The motion detection device is equipped with at least: a frame memory that delays the frame of the image input signal; a subtractor that obtains the difference between the image input signal and the frame delay signal read out from the frame memory; and the output of the subtractor and an arbitrary threshold The comparison section for comparison; the first line memory that delays the signal output by the comparison section; the second line memory that delays the signal of the first line memory again; the horizontal direction is centered on each pixel of the image signal As a block with a plurality of pixels adjacent in the vertical direction, the output from the comparison unit and the output from the first line memory and the output from the second line memory are calculated, and the sum of each pixel in the block is calculated, and the A block discrimination unit that compares with an arbitrary threshold.

The cancellation device is equipped with at least: the above-mentioned motion detection device; the output of the subtractor is multiplied by the coefficient, and the image input signal is added and subtracted, thereby removing the noise from the noise cancellation processing part; Output the third line memory that only delays the line delay of the comparison part and the singular point removal part; the fourth line memory that delays the image input signal only the line delay of the comparison part and the singular point removal part; selects the third line according to the motion detection result After the output of the line memory and the output of the fourth line memory, it is a selector that outputs the image output signal.

The image output signal selected by the selector is input to the frame memory.

Description of the drawings

FIG. 1 shows a movement detection device using block discrimination in Embodiment 1 of the present invention.

Fig. 2 shows a noise canceling device in the second embodiment of the present invention.

Fig. 3 shows a noise canceling device in Embodiment 3 of the present invention.

4A to 4C are explanatory diagrams of motion detection based on block discrimination.

Fig. 5 shows a movement detection device in Embodiment 5 of the present invention.

6A to 6E are explanatory diagrams of movement detection in Embodiment 4 of the present invention.

Fig. 7 shows a movement detection device in the fifth embodiment of the present invention.

Fig. 8 is an explanatory diagram of singularity removal in the fifth embodiment of the present invention.

Fig. 9 shows a noise canceling device in the sixth embodiment of the present invention.

Fig. 10 shows a noise canceling device in the seventh embodiment of the present invention.

11A and 11B are explanatory diagrams of gain control in Embodiment 7 of the present invention.

Fig. 12 shows a noise canceling device in the eighth embodiment of the present invention.

FIG. 13 is an explanatory diagram of level adjustment in Embodiment 8 of the present invention.

Fig. 14 shows a noise canceling device in the ninth embodiment of the present invention.

15A and 15B are explanatory diagrams of gain control in Embodiment 9 of the present invention.

Fig. 16 shows a noise canceling device in the tenth embodiment of the present invention.

Fig. 17 is an explanatory diagram of a calculation method of the mixing coefficient in the tenth embodiment of the present invention.

Fig. 18A and Fig. 18B show a frame loop type noise canceling device in a conventional example.

detailed description

The present invention provides: a movement detection device that solves the problems in the conventional movement detection device and noise cancellation device, and a noise cancellation device using the same.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.

(Embodiment 1) First, the details of (Embodiment 1) will be described.

Fig. 1 is a configuration diagram of a movement detection device in the present invention (Embodiment 1).

The frame memory 101 performs frame delay on the image input signal 108, and the subtractor 102 obtains the difference between the image input signal 108 and the frame delay signal read from the frame memory 101. The comparison unit 103 compares the output from the subtractor 102 with a predetermined threshold, and generates a desired signal based on the result. The first line memory (denoted as line memory in FIG. 1) 104 delays the signal output from the comparison unit 103 by one line (1 horizontal scanning interval), and the second line memory (denoted as line memory in FIG. 1) 105 then The output of the first line memory 104 is delayed by one line. The block determination unit 106 obtains the sum of the pixels in the block, compares the sum with a predetermined threshold, and outputs a desired signal as the value of the center pixel of the block based on the result. This block is composed of each pixel as the center, and some pixels in the vicinity of the horizontal and vertical directions. The pixels constituting the block are obtained by the comparison unit 103, the first line memory 104 and the second line memory 105. The singular point removing unit 107 removes singular points that have occurred in a series of processing by the block discriminating unit 106 and the like, and then outputs the result as a movement detection result. The process of removing singular points in the singular point removal unit 107 is to either expand the result of the block discrimination unit 106 in the horizontal and vertical directions and apply it to several pixels, or expand it to the time axis direction, or in each pixel with the The signals before processing in the block discrimination unit 106 are compared and executed.

Below, give a specific example to explain its action.

The subtractor 102 obtains a frame difference value between a signal in which the image input signal 108 is delayed by one frame in the frame memory 101 and the image input signal 108, and inputs the frame difference value to the comparison unit 103. The comparison unit 103 compares the level of the frame difference value with a predetermined threshold value, and outputs a desired signal based on the comparison result. The comparing unit 103 obtains the absolute value of the frame difference, and compares the magnitude relationship between it and the threshold. If the absolute value of the frame difference is smaller than the threshold, it outputs a 0 signal; if it is larger than the threshold, it outputs a 1-bit signal of 1.

FIG. 4A shows an example of the characteristics of the comparison unit 103. In FIG. 4A, the horizontal axis 411 represents the aforementioned frame difference; the vertical axis 412 represents the absolute value of the frame difference; straight lines 413 and 414 represent the relationship between the frame difference and the absolute value of the frame difference. In addition, the frame difference value and the absolute value of the frame difference value are divided into a region 416 and a region 417 with the threshold 415 as a boundary. The comparison unit 103 outputs 0 in the area 416 and 1 in the area 417.

At this time, low-pass filtering of the frame difference signal can also remove some noise. This 1-bit signal and the signal delayed by one line by the first line memory 104 and the output signal of the first line memory 104 delayed by the second line memory 104 by one line are input to the block discrimination unit 106.

The operation of the block determination unit 106 will be described with reference to FIGS. 4B and 4C.

In FIG. 4B, arrow 421 represents the horizontal direction of the screen, and arrow 422 represents the vertical direction of the screen. A total of 11 pixels of 3 pixels, 5 pixels, and 3 pixels are processed as one block with the pixel P6 as the center. At this time, the output of the comparison unit 103 is the pixel P9, the pixel P10, and the pixel P11, the output of the first line of memory 104 is the pixel P4, the pixel P5, the pixel P6, the pixel P7, the pixel P8, and the output of the second line of the memory 105 is the pixel P1 , Pixel P2, pixel P3.

FIG. 4C shows the output of the comparison unit 103 for each pixel in FIG. 4B. Pixels that are shaded indicate 0, and pixels that are not shaded indicate 1. In the block discrimination unit 106, for these 11 pixels, the sum of the 1-bit signals (1 or 0) output from the comparison unit 103 and the line memory 104 and the line memory 105 is obtained, and the sum is compared with a predetermined threshold. Its size relationship. At this time, the minimum value of the sum is 0 and the maximum value is 11. The block discrimination unit 106 outputs 0 as the value of the central pixel P6 if it is smaller than the threshold, and outputs a 1-bit signal of 1 as the value of the central pixel P6 if it is greater than the threshold. Here, pixels with a value of 1 are regarded as moving pixels. In FIG. 4C, the total is 7, and because the threshold is 6, the block determination unit 106 determines that P6 is a moving part and outputs 1. This processing utilizes the randomness that the moving part of the image appears uniformly in multiple adjacent pixels to a certain extent, and the randomness that noise appears scattered among adjacent pixels. The block determination unit 106 performs determination based on such a block for each pixel, and inputs the determination result to the singular point removal unit 107.

Here, lets take an example where the result of the block discrimination unit 106 is expanded and applied to several pixels in the horizontal or vertical direction, or the result of the time axis is expanded and applied. For example, the so-called enlargement and application of one pixel in the horizontal direction means that if the pixel P6 is 1 (moved) in FIG. 4B, the pixel P5 and the pixel P7 are also regarded as 1, and are regarded as moving pixels. This is because the boundary between the moving part and the non-moving part in the above-mentioned block discrimination is likely to cause misjudgment (the part that is originally 1 is judged as 0), so the animation part should be enlarged to correct this misjudgment. In this way, the signal corrected for the erroneous detection is output as the result output signal of the movement detection. By using the discrimination based on such blocks, the influence of noise contained in the input signal can be suppressed, and only the moving part of the image can be detected.

(Embodiment 2) Next, to which the present invention Ming (Embodiment 2), details will be described.

Fig. 2 is a configuration diagram of a noise canceling device in the present invention (Embodiment 2). Frame memory 201, subtractor 202, comparison unit 203, first line memory (denoted as line memory in FIG. 2) 204, second line memory (denoted as line memory in FIG. 2) 205, block discrimination unit in FIG. 2 206. The singular point removal unit 207 has the same functions as the frame memory 101, the comparison unit 103, the first line memory 104, the second line memory 105, the block discrimination unit 106, and the singular point removal unit 107 in FIG. 1. Therefore, detailed descriptions of these parts are omitted. However, the frame memory 201 performs frame delay on the image output signal of this noise canceling device.

The noise removal processing unit 208 multiplies the output of the subtractor 202 by a predetermined coefficient, and performs addition and subtraction operations with the image input signal 212, thereby removing noise. The third line memory (denoted as line memory in FIG. 2) 209 delays the output from the noise removal processing unit 208 by the line delay amount between the comparison unit 203 and the singular point removal unit 207. The fourth line memory (denoted as line memory in FIG. 2) 210 delays the image input signal 212 by the line delay amount between the comparison unit 203 and the singular point removal unit 207. The selector 211 selects the output from the noise removal processing unit 208 delayed by the third line memory 209 and the image input signal 212 delayed by the fourth line memory 210 based on the motion detection result from the singular point removal unit 207 . Then, the selector 211 outputs the selected signal as the image output signal 213.

Below, give a specific example to explain its action.

The subtractor 202 obtains the frame difference between the image output signal 213 frame-delayed by the frame memory 201 and the image input signal 212. The frame difference is used for motion detection in the same manner as the above (Embodiment 1), and the singular point removal unit 207 serves as the motion detection result signal, and outputs 1 for each pixel if it is a moving part, and outputs 0 otherwise. On the other hand, the noise removal processing unit 208 multiplies the output result (frame difference) of the subtractor 202 by a predetermined coefficient (greater than 0 and less than 1) corresponding to the difference, and inputs the signal 212 from the image according to the sign of the difference. The result is added or subtracted, thereby performing noise removal. The third line memory 209 delays the output from the noise removal processing unit 208 by the line delay amount between the comparison unit 203 and the singular point removal unit 207. In addition, the fourth line memory 210 delays the image input signal 212 by the line delay amount between the comparison unit 203 and the singular point removal unit 207. That is, in this case, as shown in FIG. 4B, a three-line block determination is performed and the center pixel corresponds to a one-line delay signal. Therefore, one-line delay is performed in the third line memory 209 and the fourth line memory 210. The selector 211 selects the output from the noise removal processing unit 208 delayed by the third line memory 209 and the image input signal 212 delayed by the fourth line memory 210 based on the motion detection result from the singular point removal unit 207 . Specifically, if the motion detection result is 1 (movement), the image input signal delayed by the fourth line memory 210 is selected; if it is 0, the image input signal delayed by the third line memory 209 is selected from the noise canceling processing section 208 The signal is output as an image output signal 213. In this noise cancellation device, since the movement detection device shown in FIG. 1 is used to complete the movement detection that suppresses the influence of noise, the noise cancellation processing unit 208 can obtain a greater noise removal effect than the conventional one, and also Can suppress the drawbacks of afterimages caused by moving parts.

(Embodiment 3)

Hereinafter, the details of (Embodiment 3) of the present invention will be described.

Fig. 3 is a configuration example of a noise canceling device in (Embodiment 3). The frame memory 301, the subtractor 302, the comparison unit 303, the first line memory (denoted as the line memory in FIG. 3) 304, the second line memory (denoted as the line memory in FIG. 3) 305, the block discrimination unit in FIG. 3 306, singular point removal unit 307, noise removal processing unit 308, third line memory (denoted as line memory in FIG. 3) 309, fourth line memory (denoted as line memory in FIG. 3) 310, selector 311, 2 and the frame memory 201, subtractor 202, comparison unit 203, first line memory 204, second line memory 205, block discrimination unit 206, singular point removal unit 207, noise removal processing unit 208, and third line memory 209 The fourth line memory 210 and the selector 211 each have the same function. Therefore, detailed descriptions of these parts are omitted. The filter 312 (denoted as LPE in FIG. 3) relies on a low-pass filtering process for the image input signal 313 mixed with noise to reduce noise components. In the noise canceling device described in FIG. 2, the amount of noise between the part of the image input signal 212 detected as a moving part and containing a lot of noise and the part where the noise is removed by the noise canceling processing unit 208 may be output as it is. Poor, resulting in an unnatural state on the surface. In this regard, by adopting the structure of FIG. 3, it is possible to reduce the apparently unnatural state caused by the above-mentioned difference in the amount of noise.

Furthermore, when the amount of noise is not that much, in order to avoid the degradation of the image quality caused by the filtering process, it can also be controlled not to perform this filtering process.

(Embodiment 4) Next, the details of (Embodiment 4) of the present invention will be described.

Fig. 5 is a configuration diagram of a movement detection device in (Embodiment 4). The image input signal 551, subtractor 502, first line memory (denoted as line memory in FIG. 5) 504, second line memory (denoted as line memory in FIG. 5) 505, and block discrimination unit 508 of FIG. 5 are It is the same as the image input signal 108, the subtractor 102, the first line memory 104, the second line memory 105, and the block determination unit 106 in FIG. Therefore, detailed descriptions of these parts are omitted.

The memory 501 corresponds to the frame memory 101 of FIG. 1. The weighting processing unit 503 corresponds to the comparator 103 in FIG. 1. The weighting processing unit 503 has more thresholds than the comparator 103 of FIG. 1 and the number of output modes is also larger. That is, the weighting processing unit 503 weights the output of the subtractor 502 in multiple levels by a predetermined threshold value. The first adder (denoted as an adder in FIG. 5) 506 and the movement coefficient generation unit 507 constitute a block discrimination unit 508.

The first adder 506 processes a plurality of pixels adjacent in the horizontal and vertical directions as one block with each pixel of the image signal as the center. The output of the weighting processing unit 503 and the output of the first line memory 504 and the second The output of the line memory 505 is the sum of the pixels in the block. The movement coefficient generation unit 507 generates a movement coefficient 532 for the center pixel of the block using a nonlinear continuous function based on the addition result of the first adder 506, and outputs the movement coefficient 532.

Below, give a specific example to explain its action.

The weighting processing unit 503 weights the output of the subtractor 502 in multiple levels by a predetermined threshold value. Here, an example in which weighting is performed in 4 levels is given for description.

FIG. 6A shows the input-output relationship of the weighting processing unit 503 that performs weighting in four levels. In FIG. 6A, the horizontal axis 611 represents the aforementioned frame difference; the vertical axis 612 represents the absolute value of the frame difference; straight lines 613 and 614 represent the relationship between the frame difference and the absolute value of the frame difference. In addition, the frame difference value and the absolute value of the frame difference value are divided into the area 618 and the area 619, and the area 620 and the area 621 with the thresholds 615, 616, and 617 as the boundary. The weighting processing unit 503 obtains the absolute value of the frame difference, and weights it in four levels from 0 to 3 with three threshold values (th1, th2, and th3) according to its size. If the absolute value is less than th1 equivalent area 618, so output 0, if it is greater than th1 and less than th2 equivalent area 619, so output 1, if it is greater than th2 and less than th3 equivalent area 620, so output 2, if it is greater than th3 equivalent area 621 so Output 3. In this way, the weighting processing unit 503 outputs a 2-bit signal representing one of 0 to 3. At this time, low-pass filtering of the frame difference signal can also remove some noise.

This 2-bit signal, the signal obtained by delaying the signal from the first line memory 504 by one line, and the signal obtained by delaying the output of the first line memory 504 by the second line memory 505 by one line further, are input to the first adder 506 . For example, the first adder 506 performs addition processing using a total of 11 pixels centered on the pixel of the pixel P6 as one block as shown in FIG. 6B. The output of the weighting processing unit 503 is pixel P9, pixel P10, and pixel P11. The output of the first line memory 504 is pixel P4, pixel P5, pixel P6, pixel P7, and pixel P8. The output of the second line memory 505 is pixel P1, Pixel P2, pixel P3. Arrow 631 and arrow 632 are the same as arrow 421 and arrow 422 in FIG. 4B, respectively. The first adder 506 performs addition processing on the weighted 2-bit values of these 11 pixels from the pixel P1 to the pixel P11 and outputs it as the value of the center pixel P6. That is, the minimum value of the addition operation result is 0, and the maximum value is 33. The larger the value, the greater the possibility that the area is a moving part. The movement coefficient generation unit 507 generates a movement coefficient 532 for the center pixel P6 using a nonlinear continuous function based on the result of the addition, and outputs FIG. 6C, which shows an example of the characteristics of the movement coefficient generation unit 507. In FIG. 6C, the horizontal axis 651 represents the addition result of the first adder 506; the vertical axis 652 represents the movement coefficient output by the movement coefficient generation unit 507; the line 653 represents the input addition result and the output movement coefficient Relationship between. As shown in FIG. 6C, the smaller the addition result, the closer the movement coefficient is to 1; the larger the addition result, the closer the movement coefficient is to 0. That is, the value of the approximate still image close to 1, and the animation with a large approximate addition result close to 0 is calculated as the movement coefficient. In the case of actual digital processing, such as outputting an integer value from 32 to 0, divide by 32 at the end of the processing. At this time, the part where the result of the addition is close to 0 (the result of the addition is less than a) is likely to be a value due to noise, so it is a static picture, that is, the movement coefficient is set to 1. In addition, when the result of the addition operation becomes larger (the part where the result of the addition operation is larger than b), there is a high possibility of animation. Therefore, it is an animation, that is, the movement coefficient is set to zero. For the remaining part, a higher-order function is used to calculate the nonlinear movement coefficient. At this time, the result of the addition operation is made so that the positions of a and b are 1 and 0, respectively.

Use Figure 6D and Figure 6E to illustrate this example. In FIG. 6D, the horizontal axis 661 represents the addition operation result; the vertical axis 662 represents the operation result; and the line 663 represents the relationship between the input addition operation result and the operation result. In addition, in FIG. 6E, the horizontal axis 671 represents the addition result; the vertical axis 672 represents the movement coefficient; and the line 673 represents the relationship between the input addition result and the output movement coefficient. First, subtract a (an integer greater than or equal to 1) from the result of the addition, and the result is greater than (-a) and less than (33-a). Secondly, this result is subtracted from 32, and it becomes greater than (a-1) and less than (32+a). If it is limited by 32, it becomes like line 663. If the result of this operation is x and a is a 4th order function of (x4)/(323) obtained from 3 pairs of x, the movement coefficient shown in Fig. 6E is obtained.

Here, the movement coefficient is expanded by 32 times (in the case of multiplying this movement coefficient, after multiplying, it must be divided by 32). If you consider digital processing, in fact, the decimal point is rounded off, so the addition result is above 22 and the shift coefficient becomes 0. As a result, more movement coefficients can be obtained using only the threshold value used in the weighting process, so processing using more continuous movement coefficients can be performed. In the method of judging the movement coefficient by the threshold value, there will be such a problem that in order to obtain a continuous coefficient, more threshold values are required, so the control becomes complicated. The present embodiment is a method of performing weighting processing, adding the weighting values in a predetermined block unit, and then obtaining the movement coefficient from a non-linear function. In this method, a smaller-scale circuit can be used to obtain a continuous movement coefficient of non-linear characteristics through weighting processing. Of course, without performing the weighting process, adding the frame difference value in a block unit as it is, and from the result, it is also possible to obtain a motion coefficient having a nonlinear characteristic (equivalent to an increase in the weighting threshold). In addition, the purpose of the addition processing in block units is to take advantage of the fact that the moving part of the image appears uniformly in multiple adjacent pixels to a certain extent, and the fact that noise has the randomness of scattered appearance between adjacent pixels . In this way, it is possible to distinguish between noise and movement to suppress the influence of noise contained in the input signal and to detect only the moving part of the image.

(Embodiment 5) Next, the details of (Embodiment 5) of the present invention will be described.

Fig. 7 is a configuration diagram of a movement detection device in (Embodiment 5). In FIG. 7, image input signal 732, memory 701, subtractor 702, weighting processing unit 703, first line memory (denoted as line memory in FIG. 7) 704, and second line memory (denoted as line memory in FIG. 7) The memory) 705, the block determination unit 731, the first adder (denoted as the adder in FIG. 7) 706, the movement coefficient generation unit 707, and the movement coefficient 733 are the same as the image input signal 551, the memory 501, and the subtractor in FIG. 5, respectively. 502, the weighting processing unit 503, the first line memory 504, the second line memory 505, the block determination unit 508, the first adder 506, the movement coefficient generation unit 507, and the movement coefficient 532 are the same. Therefore, detailed descriptions of these parts are omitted.

The second adder (denoted as an adder in FIG. 7) 708 obtains the sum of pixels in the same row in the above-mentioned block as the processing unit for the output from the weighting processing unit 703. The third adder (denoted as an adder in FIG. 7) 709 obtains the sum of the pixels in the same row in the block with respect to the output from the second row memory 705. The singular point removal unit 710 corrects the movement coefficient output by the movement coefficient generation unit 707 based on the addition result of the second adder 708 and the addition of the third adder 709, that is, if. That is, the singular point removing unit 710 either expands the moving part in the vertical direction with respect to the movement coefficient output by the movement coefficient generating unit 707, or expands the moving part in the horizontal direction with respect to the movement coefficient from the movement coefficient generating unit 707, thereby Remove the singularities that are likely to occur in the block-by-block processing.

Below, give a specific example to explain its action.

The movement coefficient generator 707 outputs the movement coefficient for the image input signal 732 in the same method as (Embodiment 4). The second adder 708 adds two-bit values weighted by the weighting processing unit 703 for the three pixels of the pixel P9, the pixel P10, and the pixel P11 in FIG. 6B. In addition, the third adder 709 adds the weighted 2-bit values for the three pixels of the pixel P1, the pixel P2, and the pixel P3 in FIG. 6B output from the second line memory 705. The respective addition results are input to the singular point removal unit 710.

The singular point removal unit 710 corrects the movement coefficient from the movement coefficient generation unit 707 according to the result of the addition or enlarges the movement part in the vertical direction, or expands the movement part in the horizontal direction for the movement coefficient from the movement coefficient generation unit 707 to remove Singularities that easily occur in processing in units of blocks.

Next, an example of processing in the singular point removal unit 710 is shown. First, for each pixel with respect to the movement coefficient shown in FIG. 6E shown in (Embodiment 4) output by the movement coefficient generating unit 707, the minimum value of 1 to 2 pixels on the left and right is taken as the movement coefficient of the center pixel. As a result, the moving part can be enlarged in the horizontal direction. In addition, in the expansion of the moving part in the vertical direction, as shown in FIG. 8, the result of the addition (the minimum value is 0, the maximum value is 9) of the second adder 708 or the third adder 709 corrects the center pixel of the processing block. The movement coefficient of P6. The arrow 801 and the arrow 802 in FIG. 8 from the pixel P1 to the pixel P11 correspond to the arrow 601 and the arrow 602 in FIG. 6 from the pixel P1 to the pixel P11. The singular point removal unit 710, for example, if the result of the addition transportation is 7 or more, makes the movement coefficient of the pixel P6 0 to correct the movement coefficient expansion portion from the movement coefficient generation unit 707. Through such processing, it is possible to remove singularities that are likely to occur in processing in units of blocks.

(Embodiment 6) Next, the details of (Embodiment 6) of the present invention will be described.

Fig. 9 is a configuration example of a noise canceling device in (Embodiment 6). In FIG. 9, the image input signal 951, memory 901, subtractor 902, weighting processing unit 903, the first line memory (denoted as line memory in FIG. 9) 904, the second line memory (denoted as line memory in FIG. 9 (Memory) 905, block discrimination unit 931, first adder (denoted as adder in FIG. 9) 906, each block of movement coefficient generation unit 907 and the image input signal 551 of FIG. 5, memory 501, subtractor 502, The blocks of the weighting processing unit 503, the first line memory 504, the second line memory 505, the block determination unit 508, the first adder 506, and the movement coefficient generation unit 507 have the same function. In addition, the blocks of the second adder (denoted as the adder in FIG. 9) 908, the third adder (denoted as the adder in FIG. 9) 909, and the singular point removal unit 910 are the same as the second adder in FIG. Each block of the block 708, the third adder 709, and the singular point removal unit 710 has the same function. Therefore, detailed descriptions of these parts are omitted.

The fifth line memory (denoted as line memory in FIG. 9) 911 delays the image input signal 951 by one line. The sixth line memory (denoted as line memory in FIG. 5) 912 delays the signal read from the memory 901 by one line. The second subtractor 913 obtains the difference between the signal from the fifth line memory 911 and the signal from the sixth line memory 912. The first multiplier 914 multiplies the output of the second subtractor 913 by a predetermined gain (more than 0 and less than 1). The second multiplier 915 multiplies the result of the first multiplier 914 by the movement coefficient determined by the movement coefficient generation unit 907 and the singular point removal unit 910. The adder-subtractor 916 adds and subtracts the output of the second multiplier 915 and the signal of the fifth line memory 911 based on the sign of the subtraction result in the second subtractor 913. The memory 901 outputs a signal in which the image output signal 952 from the adder-subtractor 916 is delayed (1 frame-1 line). This is a cyclic noise canceling device using the movement detection device shown in (Embodiment 4) or (Embodiment 5).

Below, give a specific example to explain its action.

The subtractor 902 obtains the difference between the image output signal 952 (the signal from which the noise has been removed) delayed by (1 frame-1 line) in the memory 901 and the image input signal 951. The reason why the delay amount in the memory 901 is one line less than one frame is that the output of the motion coefficient in the motion detection device shown in (Embodiment 4) or (Embodiment 5) lags behind the image input signal 951 by one line. With respect to this difference value, according to the method described in (Embodiment 4) or (Embodiment 5), the movement coefficient is determined by the movement coefficient generation unit 907 and the singular point removal unit 910. On the other hand, the second subtractor 913 finds that the image input signal 951 is delayed by one line in the fifth line memory 911 and the signal from the memory 911 is delayed by one line in the sixth line memory 912. The difference between the signals. The first multiplier 914 multiplies the difference by an arbitrary gain (greater than 0 and less than 1) 917. The gain 917 is a coefficient that is multiplied independently by the movement coefficient, and is one of the elements that determines the amount of circulation. If the gain 917 is increased, the noise removal effect becomes greater. When digital processing is assumed, the first multiplier 914 is actually a 32-fold expansion device, that is, after multiplying the difference by an integer value greater than 0 and less than 32, the result is divided by 32. The second multiplier 915 multiplies the result of the first multiplier 914 by the movement coefficient determined by the movement coefficient generation unit 907 and the singular point removal unit 910. As this movement coefficient, for example, the coefficient shown in FIG. 6E shown in (Embodiment 4) is multiplied. That is, the second multiplier 915 multiplies the integer value from 0 to 32 according to the movement, and then divides by 32. Thus, the gain (circulation amount) that is the same for all pixels multiplied by the first multiplier 914 is adjusted based on the amount of movement of the pixel. With respect to the signal of the fifth line of memory 911, the result of the second multiplier 915 is added and subtracted in the adder-subtractor 916 based on the sign of the subtraction result in the second subtractor 913, whereby noise can be removed.

This noise canceling device uses the movement detection device described in (Embodiment 4) or (Embodiment 5). By using these motion detection devices, motion detection from strong electric field to weak electric field can be completed to suppress the influence of noise, and at the same time, a more continuous motion coefficient can be obtained between adjacent pixels according to the amount of image movement. Therefore, the noise canceling device in this embodiment can increase the noise canceling effect, and at the same time, can reduce the superficial unnatural state of the image that is caused by afterimages and processing using discontinuous coefficients.

(Embodiment 7) Next, the details of (Embodiment 7) of the present invention will be described.

Fig. 10 is a configuration example of a noise canceling device in (Embodiment 7). In FIG. 10, the image input signal 1051, the memory 1001, the subtractor 1002, the weighting processing unit 1003, the first line memory (denoted as line memory in FIG. 10) 1004, and the second line memory (denoted as line memory in FIG. 10) Memory) 1005, block determination unit 1031, first adder (denoted as adder in FIG. 10) 1006, and each block of movement coefficient generation unit 1007, and the image input signal 551, memory 501, and subtractor 502 of FIG. 5 The blocks of the weighting processing unit 503, the first line memory 504, the second line memory 505, the block discrimination unit 508, the first adder 506, and the movement coefficient generation unit 507 have the same function. In addition, the blocks of the second adder (denoted as the adder in FIG. 10) 1008, the third adder (denoted as the adder in FIG. 10) 1009, and the singular point removal unit 1010 are the same as the second adder in FIG. Each block of the block 708, the third adder 709, and the singular point removal unit 710 has the same function. In addition, the fifth line memory (denoted as line memory in FIG. 10) 1011, the sixth line memory (denoted as line memory in FIG. 10) 1012, the second subtractor 1013, the first multiplier 1014, and the second multiplier 1015. Each block of the adder-subtractor 1016 is the same as the fifth line memory 911, the sixth line memory 912, the second subtractor 913, the first multiplier 914, the second multiplier 915, and the adder-subtractor 916 in FIG. 9 Each box has the same function. Therefore, detailed descriptions of these parts are omitted.

The first selector 1017 is controlled by the control signal 1043 to select one of the first gain 1041 and the second gain 1042. The first selector 1017 switches a plurality of gains in accordance with a control signal 1043 determined for each pixel in the horizontal and vertical directions and the frame or field direction.

Below, give a specific example to explain its action.

In the noise canceling device that removes noise according to the method described in (Embodiment 6), the image input signal 951 is multiplied by a gain in the first multiplier 914 by the difference result in the second subtractor 913. In the noise canceling device in this embodiment, multiple gains are used and they are switched for each pixel in the horizontal and vertical directions and the frame or field direction.

11A and 11B show examples of the above-mentioned handover. The two graphs show the gain in each pixel within a frame or a field. G1 represents the first gain 1041 and G2 represents the second gain 1042. The picture 1101 and the picture 1102 are adjacent frames or fields. In the screen 1101 and the screen 1102, the horizontal direction and the vertical direction G1 and G2 are alternately switched. In addition, for pixels at the same position in the screen, G1 and G2 are also alternately switched between screen 1101 and screen 1102.

As described above, the control signal 1403 that switches the two gains between adjacent pixels in the horizontal and vertical directions and the frame or field direction occurs, and is input to the selector 1017. The first selector 1017 switches the two gains according to this control signal 1043. In this case, for one gain, the other gain is designed to be smaller than it, so that a small noise is intentionally added, which can reduce the afterimages and feel that the image in the spatial direction that is caused by the increase in gain. It looks unnatural on the surface. Furthermore, by switching the gain, the other gain can be set larger than in the case of processing with one gain, so a greater noise removal effect can be obtained. Therefore, if you want to set it so that it doesn't feel like sticking in a strong electric field, you can also set another gain to 1. In the process of noise removal, pixels with the same cyclic coefficient are generally easy to gather (especially for areas close to a still picture), so it is easy to feel that the image appears to be pasted and looks unnatural on the surface. However, by controlling the gain in pixel units in this way, this phenomenon can be suppressed. Of course, this method can also be applied to other noise removal devices.

(Embodiment 8) Next, the details of (Embodiment 8) of the present invention will be described.

Fig. 12 is a configuration example of a noise canceling device in (Embodiment 8). In FIG. 12, the image input signal 1251, memory 1201, subtractor 1202, weighting processing unit 1203, first line memory (denoted as line memory in FIG. 12) 1204, and second line memory (denoted as line memory in FIG. 12) Memory) 1205, block determination unit 1231, first adder (denoted as adder in FIG. 12) 1206, and each block of movement coefficient generation unit 1207, and the image input signal 551, memory 501, and subtractor 502 of FIG. 5 The blocks of the weighting processing unit 503, the first line memory 504, the second line memory 505, the block discrimination unit 508, the first adder 506, and the movement coefficient generation unit 507 have the same function. In addition, the blocks of the second adder (denoted as the adder in FIG. 12) 1208, the third adder (denoted as the adder in FIG. 12) 1209, and the singular point removal unit 1210 are the same as those of the second adder in FIG. The blocks of the adder 708, the third adder 709, and the singular point removal unit 710 have the same function. In addition, the fifth line memory (denoted as line memory in FIG. 12) 1211, the sixth line memory (denoted as line memory in FIG. 12) 1212, the second subtractor 1213, the first multiplier 1214, and the second multiplier 1215. Each block of the adder-subtractor 1216 is the same as the fifth line memory 911, the sixth line memory 912, the second subtractor 913, the first multiplier 914, the second multiplier 915, and the adder-subtractor 916 in FIG. 9 Each box has the same function. In addition, the first selector 1217 has the same function as the first selector 1017 in FIG. 10. The first gain 1241, the second gain 1242, and the control signal 1243 correspond to the first gain 1041, the second gain 1042, and the control signal 1043 in FIG. 10. Therefore, detailed descriptions of these parts are omitted.

If the absolute value level of the signal output from the second subtractor 1213 is equal to or lower than a predetermined level, the level adjustment unit 1218 continuously adjusts it to a value smaller than the original absolute value. The adjusted value is input to the first multiplier 1214.

Below, give a specific example to explain its action.

In the noise canceling devices shown in (Embodiment 6) and (Embodiment 7), the first multiplier 1214 multiplies the difference result in the second subtractor 1213 by the gain of the image input signal 1251. In this embodiment, if the absolute value of the difference is equal to or less than the predetermined value, the level adjustment unit 1218 continuously adjusts the absolute value of the difference to a value smaller than the original absolute value.

FIG. 13 shows an example of the input and output characteristics of the level adjustment unit 1218. In FIG. 13, the horizontal axis 1301 represents the difference value input to the level adjustment unit 1218; the vertical axis 1302 represents the value output by the level adjustment unit 1218; and the broken line 1303 represents the input/output relationship of the level adjustment unit 1218. Let the difference input to the level adjustment unit 1218 be D and the threshold be th. The level adjustment unit 1218 outputs D/2 if the input difference D is th/2 or less, outputs 3D/2-th/2 if it is greater than th/2 and less than th, and outputs D if it is greater than th. According to this input/output characteristic, the noise removal effect becomes small in the difference value below the threshold th. This is intended to limit the amount of low-level noise removal to a certain extent. However, in the noise removal process at the time of a weak electric field, it is easy to produce an unnatural state on the surface, such as a film on the entire screen. As in the cost embodiment, low-level noise remains to a certain extent, thereby reducing the apparently unnatural state. The threshold value is controlled according to the strength level of the electric field, and the level adjustment can be made invalid when the strong electric field is changed, and the low-level noise removal amount can be restricted as the weak electric field is changed. Since the threshold value of the weighting processing unit 1203 is increased in the case of a weak electric field, there is a tendency to increase the feeling that a film is attached to the entire image accompanying the noise removal processing. However, by performing such level adjustment, it is possible to suppress the unnatural state on the surface, and also to focus on removing noise with a high level that is more conspicuous. Instead of adjusting the difference value, several level adjustment methods such as reducing the gain below the threshold value can be considered, but the same point is that the noise removal effect of a low difference level is suppressed.

(Embodiment 9) Next, the details of (Embodiment 9) of the present invention will be described.

Fig. 14 is a configuration example of a noise canceling device in (Embodiment 9). The memory 1401 in FIG. 14, the subtractor 1402, the weighting processing unit 1403, the first line memory (denoted as line memory in FIG. 14) 1404, the second line memory (denoted as line memory in FIG. 14) 1405, and the first addition The blocks of the unit (denoted as the adder in FIG. 14) 1406, the movement coefficient generation unit 1407, and the block discrimination unit 1431 are combined with the memory 501, subtractor 502, weighting processing unit 503, and first line memory 504 in FIG. The blocks of the second line memory 505, the first adder 506, the movement coefficient generation unit 507, and the block discrimination unit 508 have the same function. In addition, the blocks of the second adder (denoted as the adder in FIG. 14) 1408, the third adder (denoted as the adder in FIG. 14) 1409, and the singular point removal unit 1410 are the same as those of the second adder in FIG. The blocks of the adder 708, the third adder 709, and the singular point removal unit 710 have the same function. In addition, the fifth line memory (denoted as line memory in FIG. 14) 1411, the sixth line memory (denoted as line memory in FIG. 14) 1412, the second subtractor 1413, the first multiplier 1414, and the second multiplier 1415. Each block of the adder-subtractor 1416 is the same as the fifth line memory 911, the sixth line memory 912, the second subtractor 913, the first multiplier 914, the second multiplier 915, and the adder-subtractor 916 in FIG. 9 Each box has the same function. In addition, the first selector 1417 has the same function as the first selector 1017 in FIG. 10. In addition, the level adjustment unit 1418 has the same function as the level adjustment unit 1218 in FIG. 12. In addition, the first gain 1441, the second gain 1442, and the control signal 1443 correspond to the first gain 1041, the second gain 1042, and the control signal 1043 in FIG. 10. Therefore, detailed descriptions of these parts are omitted.

The averaging circuit 1419 obtains the sum of the pixels in the corresponding processing block shown in (Embodiment 4) for the image input signal 1451 and the signal of the fifth line memory 1411, and calculates the average signal level. The gain adjustment unit 1420 adjusts the gain multiplied by the first multiplier 1414 based on the average value calculated by the average circuit 1419.

Below, give a specific example to explain its action.

In the noise canceling device that removes noise according to the method shown in (Embodiment 8), the first multiplier 1414 multiplies the output of the level adjustment unit 1418 by the gain output by the first selector 1417. In this embodiment, this gain is controlled based on the average signal level of the pixel to be processed and the surrounding image.

15A and 15B show an example of this control. As shown in FIG. 15A, the averaging circuit 1419 calculates the pixels in the data 1502 of the image input signal 1451 and the data 1501 from the fifth line memory 1411 in the corresponding processing block shown in (Embodiment 4), which is equivalent to The average signal level of 8 pixels from the pixel P4 to the pixel P11 in FIG. 6B. Then, the gain adjustment unit 1420 adjusts the gain based on this average signal level. The gain adjustment unit 1420 relatively reduces the gain for the bright portion where the signal level is high, and increases the gain for the dark portion where the gradation level is small.

FIG. 15B shows an example of the characteristics of the gain adjustment unit 1420. In FIG. 15B, the horizontal axis 1524 represents the average signal level from the averaging circuit 1419; the vertical axis 1521 represents the gain adjustment amount in the gain adjustment unit 1420; the broken line 1523 represents the average signal level and the gain adjustment amount. An example of a relationship. Here, the gain adjustment unit 1420 adjusts the gain from the first selector 1417 in a broken line as shown in FIG. 15B, for example, and sets the gain to approximately 90% above a certain average signal level th. That is, the initial gain setting is set based on the case where the average signal level is small. By adjusting the gain from the average signal level in this way, the noise removal effect can be improved for each pixel in a region where the amount of noise tends to increase in a small signal level in the same frame and in the same field. In addition, there is no need to link with the AGC circuit. In addition, here, the average signal level is calculated by taking such distorted 8 pixels as one block from the perspective of the circuit and the amount of calculation, but this block can be arbitrarily determined. The key is to adjust the gain after obtaining the average signal level of multiple pixels. In the method of using the input signal level itself, the control level is easily changed in the spatial direction due to the influence of noise. In this embodiment, in order to reduce this variation, control is performed by averaging values based on surrounding pixels.

(Embodiment 10) Next, the details of (Embodiment 10) of the present invention will be described.

Fig. 16 is a configuration example of a noise canceling device in (Embodiment 10). 16 memory 1601, subtractor 1602, weighting processing unit 1603, first line memory (denoted as line memory in FIG. 16) 1604, second line memory (denoted as line memory in FIG. 16) 1605, first addition The blocks of the block (denoted as the adder in FIG. 16) 1606, the movement coefficient generation unit 1607, and the block discrimination unit 1631 are the same as the memory 501, the subtractor 502, the weighting processing unit 503, and the first line memory 504 in FIG. The blocks of the second line memory 505, the first adder 506, the movement coefficient generation unit 507, and the block discrimination unit 508 have the same function. In addition, the blocks of the second adder (denoted as the adder in FIG. 16) 1608, the third adder (denoted as the adder in FIG. 16) 1609, and the singular point removal unit 1610 are the same as those of the second adder in FIG. The blocks of the adder 708, the third adder 709, and the singular point removal unit 710 have the same function. In addition, the fifth line memory (denoted as line memory in FIG. 16) 1611, the sixth line memory (denoted as line memory in FIG. 16) 1612, the second subtractor 1613, the first multiplier 1614, and the second multiplier 1615. Each block of the adder-subtractor 1616 is the same as the fifth line memory 911, the sixth line memory 912, the second subtractor 913, the first multiplier 914, the second multiplier 915, and the adder-subtractor in FIG. 9 Each block of 916 has the same function. In addition, the first selector 1617 has the same function as the first selector 1017 in FIG. 10. In addition, the level adjustment unit 1618 has the same function as the level adjustment unit 1218 in FIG. 12. In addition, the average circuit 1619 and the gain adjustment unit 1620 have the same functions as the average circuit 1419 and the gain adjustment unit 1420 in FIG. 14. Therefore, detailed descriptions of these parts are omitted.

The filter processing unit 1621 performs spatial low-pass filter processing on the output of the adder-subtractor 1616. The mixing coefficient calculation unit 1622 determines the mixing ratio between the output signal of the filter processing unit 1621 and the output signal of the adder-subtractor 1616 based on the continuous movement coefficient values obtained by the movement coefficient generation unit 1607 and the singular point removal unit 1610 . The mixing processing unit 1623 mixes the output signal of the filter processing unit 1621 and the output signal of the adder-subtractor 1616 based on the mixing coefficient determined by the mixing coefficient calculation unit 1622. The contour detection unit 1624 extracts the contour part of the image from the output signal of the adder-subtractor 1616. The second selector 1625 switches between the output of the mixing processing unit 1623 and the output signal of the adder-subtractor 1616 in accordance with the output from the contour detection unit 1624.

Below, give a specific example to explain its action.

In this embodiment, regarding the image input signal 1651, regarding the output signal of the noise canceling device whose noise is removed by the methods shown in (Embodiment 6) and (Embodiment 7), (Embodiment 8) and (Embodiment 9), The filter processing unit 1621 performs low-pass filter processing in the spatial direction. For example, the filter processing unit 1621 in FIG. 16 does not add a line memory and uses only a filter in the horizontal direction. By mixing the output signal of the filter processing unit 1621 and the input signal of the filter processing unit 1621, the effect of the spatial filter is adjusted. At this time, the mixing coefficient calculation unit 1622 uses the value of the movement coefficient obtained by the movement coefficient generation unit 1607 and the singular point removal unit 1610 to determine the mixing coefficient. That is, in the noise canceling devices shown in (Embodiment 6) and (Embodiment 7), (Embodiment 8), and (Embodiment 9), since the effects of afterimages and the like are taken into consideration, the animation portion where noise is difficult to remove may sometimes Residual noise will occur. As a result, sometimes an unnatural state on the surface may occur. In this embodiment, this unnatural state is to be reduced. For this reason, for example, for the movement coefficient shown in FIG. 6E shown in (Embodiment 4), the mixing ratio shown in FIG. 17 is set.

This mixing ratio is explained in conjunction with Fig. 17. The horizontal axis 1701, the vertical axis 1702, and the curve 1703 in FIG. 17 correspond to the horizontal axis 671, the vertical axis 672, and the curve 673 in FIG. 6E, respectively. Therefore, detailed descriptions of these parts are omitted. As shown in Figure 17, the closer the pixel to the animation, the smaller the motion coefficient, the larger the mixing ratio of the spatial filtering. Set a certain threshold th1704, and if the movement coefficient 1702 is equal to or greater than th (nearly stationary) area 1709, the mixing ratio of the output of the filter processing unit 1621 is set to 0 (that is, the noise shown before (Embodiment 9) Eliminate the output of the device itself); if it is an area 1708 above 3th/4 and below th, set its ratio to 1/4; if it is an area 1707 above th/2 and below 3th/4, set its ratio If it is the area 1706 above th/4 and below th/2, set its ratio to 3/4; if it is the area 1705 below th/4, set its ratio to 1 (ie The output of the filter processing unit 1621 itself). By changing this threshold th1704, the filtering effect in the spatial direction can be controlled. Therefore, the threshold can be increased as the electric field becomes weak. Furthermore, several calculation methods for calculating this mixing coefficient (above-mentioned mixing ratio) using a function can also be considered. Based on the mixing coefficient determined in this way, the unfiltered signal and the filtered signal are mixed in the mixing processing unit 1623, thereby adjusting the effect of the filtering processing in the spatial direction according to the amount of movement. Furthermore, in order to prevent blurring of the image due to the addition of spatial filtering, the contour detection unit 1624 detects contour portions. For example, in the case where only filtering is applied in the horizontal direction, the horizontal difference is obtained. If the difference is above a certain level, it is judged to be a contour, and the second selector 1625 is controlled for the left and right 1 and 2 pixels so that they are not added. Filtering in the spatial direction. As a result, a signal that has not been filtered is output. Through such processing, it is possible to reduce the unnatural appearance of the surface of the animation portion caused by the residual noise that is likely to occur with the cyclic noise removal processing. At the same time, without additional contour correction circuit, it is possible to suppress the blur of the contour part caused by the filtering process in the spatial direction to a certain extent.

As described above, in the present invention, the randomness of noise is used to obtain the frame difference between the image input signal frame-delayed signal and the image input signal, and the difference is compared with a predetermined threshold. For the output signal of this result, the sum of the blocks of several pixels adjacent in the horizontal and vertical directions is obtained with each pixel as the center, and this sum is compared with a predetermined threshold value and used as the value of the center pixel of the block Output the desired signal. Furthermore, this output is either expanded to apply several pixels in the horizontal and vertical directions, or expanded to the time axis direction, or in each pixel, compared with the signal before the processing of the block discrimination unit to remove the existing block Singularities that are prone to occur in the handling of units. In this way, the influence of noise contained in the image input signal can be suppressed and only the moving part of the image can be detected. Furthermore, by multiplying the frame difference value by a desired coefficient and performing addition and subtraction operations with the image input signal, the noise-removed signal and the image input signal are selected in each pixel based on the motion detection result to form an image output signal. In this way, the noise removal effect can be increased, and at the same time, the disadvantage of the attenuation tail caused by the movement of the image can be suppressed.

In addition, in the present invention, for an image input signal mixed with noise, in addition to detecting a moving part with a large frame difference, it also distinguishes a moving part of an image with a frame difference equal to or lower than the noise level from noise and noise. Carry out movement detection that suppresses the influence of noise, and build a noise removal device using this movement detection device. In this way, compared with the conventional one, while increasing the noise removal effect, it is also possible to suppress the occurrence of the attenuation tail of the moving part that is accompanied by the increase of the noise removal effect.

In addition, according to the present invention, when there is a lot of noise, it is possible to reduce the apparently unnatural state caused by the difference in the amount of noise from the noise-removed portion.

In addition, the present invention provides: using the randomness of noise to perform movement detection based on the information of several pixels in the spatial direction, without looking up tables, etc., using a method of controlling the circuit scale, and outputting more continuous output between adjacent pixels according to the amount of image movement. Movement detection device for moving coefficients. Furthermore, the construction of a noise canceling device using the motion detection device can improve the effect of noise removal while suppressing afterimages and discontinuous coefficient processing and the appearance of the image on the surface. Unnatural state.

Furthermore, in the present invention, in the dark part where the amount of surrounding noise tends to increase, the effect of noise is suppressed without using information such as an AGC circuit to appropriately improve the noise removal effect. In addition, it reduces the apparently unnatural state caused by the residual noise caused by the noise removal, so that it has a weak electric field to a strong electric field. It improves the noise removal effect and prevents the accompanying occurrence of animation and still pictures. Afterimages and unnatural appearance on the surface.

According to the present invention, it can be provided that for the weighted frame or field difference value, the adjacent pixels are used as a processing unit to perform the addition processing, thereby, the continuous movement is output according to the nonlinear function without the threshold value. Coefficient of movement detection device. In this way, the difference between noise and movement that uses the randomness of noise can suppress the influence of noise contained in the input signal and provide a continuous movement coefficient according to the movement of the image. Furthermore, the construction of a noise canceling device using the motion detection device can increase the noise removal effect while reducing the afterimages and the surface appearance of the image produced by processing using discontinuous coefficients. The unnatural state. Furthermore, by switching multiple gains in each pixel, it is possible to deliberately add minute noises, which can reduce the afterimages and the perception that the image in the spatial direction appears to be pasted, etc. that are caused by the increase in gain. The unnatural state of getting up. In addition, it is possible to achieve an improvement in the noise removal effect brought about by increasing the gain. In addition, when the electric field is weak, the low-level difference is adjusted to intentionally limit the noise removal effect of the low-level difference level, thereby reducing the unnatural state that seems to be a layer of film on the surface, and at the same time , You can focus on removing large-level noise that is easily noticeable. In addition, the gain is adjusted based on the average signal level for a plurality of adjacent pixels, so that even in the same screen, the noise can be increased for each pixel in a region where the amount of noise is likely to increase, and the signal level is small. Remove the effect. Furthermore, based on the movement coefficient from the movement detection device and the contour information from the contour detection unit, a filter in the spatial direction is appropriately used in combination, so that the contours of the moving part can be reduced without blurring the contour of the moving part. The residual noise that is prone to noise removal processing causes an unnatural appearance on the surface of the animation part.

By using such a loop-type noise canceling device, from a weak electric field to a strong electric field, for animation and still images, it is possible to simultaneously improve the noise removal effect and reduce the accompanying afterimages and feel as if the image is pasted on the surface. The unnatural state of getting up.

In addition, the present invention can improve the noise removal effect for each pixel even in a region with a small signal level where the amount of noise is likely to increase in the same screen.

In addition, according to the present invention, without blurring the outline of the moving part, it is possible to reduce the unnatural appearance of the animation part caused by the residual noise that easily occurs with the cyclic noise removal process.

By using such a cyclic noise canceling device, it is possible to simultaneously improve the noise canceling effect for animation and still images from weak electric field to strong electric field, and prevent the accompanying afterimages and unnatural appearance on the surface.

The movement detection device of the present invention and the noise canceling device using the device can increase the noise canceling effect, and at the same time, can suppress the occurrence of attenuation tails of the moving part accompanying the increase of the noise canceling effect. Furthermore, the noise removal effect can be increased from a weak electric field to a strong electric field, and at the same time, it is possible to cancel the superficially unnatural state such as after-images and feelings that appear to be pasted that accompany it.