System and method determining spectrum using fuzzy inference algorithm employing measurement from led sensor

Description: TECHNICAL FIELD [Detailed description of the invention]

【0001】

[Technical field to which the invention belongs]

The present invention relates to determining a spectrum (spectral, frequency domain) based on a non-spectral input.

【0002】

[Conventional technology]

Automatic online color calibration systems can be very effective when used with online color measurement systems. In an online color measurement system, the spectrophotometer does not require any changes to the printer and is a copy sheet that moves within the printer without interfering with or interrupting normal printing or the movement of the print sheet along the paper path. The spectrophotometer may be attached to the paper path, preferably the output path after fusion (melt fixation) or drying, and the spectrophotometer is the exact color of the test color patch (inversion) printed on the moving sheet. Measurements are made as those sheets pass through the spectrophotometer. It allows for completely closed loop color control of the printer.

【0003】

A typical spectrophotometer provides color information from the test surface at different wavelengths of light in terms of the reflectance or transmittance of the measured light. It is desirable for this spectrophotometer to provide separate electrical signals that correspond to different levels of reflected light received from different illumination wavelength ranges or channels.

【0004】

Each different illumination wavelength ranges or pairs at different levels of the reflected light received from the channel to the known devices that can provide separate electrical signals are response, it is based on the lattice made by Ocean Opti box Ltd. Spectrophotometers, LED-based sensors marketed by "ColorSavvy" or Accuracy Microsensors, Gretag MacBeth (viptronic) ), Ex color, other spectrophotometers with X-light (Rite) are included. However, these devices are believed to have considerable cost, measurement time, target displacement error, and / or other difficulties to use in real-time printer online measurements.

【0005】

As used herein, unless otherwise indicated, the term "spectrophotometer", as broadly defined herein, is a spectrophotometer, colorimeter (hydrometer), concentration. A meter (hydrometer) may be included. The definitions or uses of the above terms may vary and may differ among different scientists and engineers. However, the following description attempts to provide some brief description related to and distinguishing between the terms "spectrophotometer", "colorimeter", and "densitometer". Because they can be used in special cases in the specification that provide components for an online color printer color correction system, they need to necessarily limit the claims. Because there is no.

【0006】

A typical "spectrophotometer" measures the reflectance of an illuminated object in question over a number of light wavelengths. Under these circumstances, conventional spectrophotometers use 16 or 32 illumination channels measuring 380 nm to 730 nm to cover the color spectrum or wavelength range visible to the human eye. A typical spectrophotometer gives color information in terms of the reflectance or transmittance of the measured light at different wavelengths of light from the test surface. (This is to more precisely measure what the human eye sees as a combined image of wide white light spectrum image reflectance, but spectrophotometers use reflected light from different illumination wavelength ranges or channels. It is desirable to give separate electrical signals that correspond to different levels of.

【0007】

A "colorimeter" usually has three illumination channels: red (red), green (green), and blue (blue). That is, in general, a "colorimeter" shall read these three values (red, green, and blue, or "RGB") by an optical sensor or detector (where these optical sensors or detectors). The vessel is a color test that is sequentially (sequentially) illuminated using a single white light lamp with three different color filters, or a red, green, blue illuminator (illuminator) such as three different color LEDs. Receives reflected light from the surface). Therefore, it should be considered different from the "spectrophotometer" in that it provides output color information in the amount of three colors (used) known as RGB, or a limited special case of the "spectrophotometer". Is.

【0008】

The amount of three colors can be used to represent colors in tricoordinate space through several types of transformations. Other RGB conversions to "device independent color space" (ie, RGB converted to traditional L * a * b) are generally known as color conversion variants, or "looks". Use the "uptable" system.

【0009】

A "densitometer" generally has only a single channel and is in the wavelength range (which may be wide or narrow) from a test surface, such as a toner test patch developed on a photoconductor. It only measures the magnitude of the reflectance of light at a given angle. A single light source (light source) such as an IR (infrared) LED, visible LED, or incandescent lamp can also be used. The output of the densitometer detector is programmed to give the optical density of the sample. This type of densitometer is basically a "color blind". For example, the cyan test patch and the magenta test patch will have the same optical density when viewed by a densitometer, but of course they are different colors.

【0010】

[Summary of Invention]

Multiple LED reflectance spectrophotometers, as in the examples of embodiments herein, are a special class of spectrophotometers that typically use narrow band (frequency band) or monochromatic light to illuminate the target. It can also be considered to belong to. Others having a broadband illumination source may be a flash type xenon lamp spectrophotometer or an incandescent lamp spectrophotometer. Spectrophotometers are typically programmed to use conversion algorithms to provide more detailed reflectance values by using four or more channel measurements (eg, 10 or more channel measurements). Has been done. It is a regular 3-channel colorimeter that cannot provide accurate, human-eye-related, reflection spectrum measurements (because those regular 3-channel colorimeters are sufficient for that purpose (only 3). It is different from) which does not have measurement).

【0011】

The printer color control system is "on line" (ie, the printed output medium is still in the sheet carrier, or the paper of the printing engine, for real time and for the use of fully automatic printer color correction. It is desirable to dynamically measure the color of the test patch on the printed output medium (while in the path).

【0012】

For low cost implementation of color sensors, multiple light emitter devices are used as illumination sources, eg, have 8, 10, 12, or 16 LEDs. Each LED is selected to have a narrowband response curve in spectral space. So, for example, 10 LEDs correspond to 10 measurements on the reflectance curve. An LED, or the equivalent of a color sensor based on other multiplex emitters, such as a laser, is turned on one at a time, for example, when the measured medium passes through the transport section of the printer. Can be switched. The reflected light is then detected by a photodetector and the corresponding voltage is integrated and normalized (normalized, standardized) using white tiles.

【0013】

A linear or cubic spline algorithm can be used to obtain a smooth spectral response similar to that of a Gretag spectrophotometer. It interpolates and extrapolates data points blindly without knowledge of color space. Unfortunately, blind extrapolation with such a small number (eg 10) causes an error due to the lack of measurements at wavelengths below 430 nm and above 660 nm (due to the lack of LEDs at these wavelengths). It may cause it.

【0014】

Although some light sources may not produce spectral content at the remote end of the visible spectrum, the devices and methods of the invention use integrated sensor measurements at special wavelengths. The reflectance value determines the fully populated wavelength spectra. Integrated multiplex illuminant measurement from a non-fully illuminant populated color sensor by using a color detection system and a reconstruction algorithm based on the spectral characteristics of the source. Is transformed into a fully populated spectral curve.

【0015】

The algorithm according to the invention utilizes a reference database containing training samples showing reflection spectra (eg, obtained via a Gretag spectrophotometer) and their corresponding LED sensor outputs. The fuzzy inference algorithm uses the non-linear model obtained from the training sample of this reference database.

【0016】

These and other objectives, advantages, and salient features of the present invention are set forth in, or are apparent from, the following description of the exemplary embodiments.

【0017】

BEST MODE FOR CARRYING OUT THE INVENTION

The spectrophotometers of the disclosed embodiments are mounted on one side of the print sheet output path of a color printer and optically evaluate the color imprinted output sheets as they pass through the spectrophotometer. A spectrophotometer that is variably separated from the sheet and does not come into contact with the sheet or interfere with the normal movement of the sheet. In particular, it is printed by the printer on the printer's actual print sheet output during regular or predetermined printer operation intervals (intervals) between regular print runs or between print jobs. Can be used to measure a large number of color test patch samples. These color test sheet print intervals may be at regular time intervals and / or at each machine "cycle up" or as instructed by the system software. Good. The spectrophotometer may be mounted on one side of the paper path of the machine, and if it is desirable to use a duplex color test sheet, two spectrophotometers may be mounted on both sides of the paper path. You may.

【0018】

Relatively frequent color calibration of color printers is highly desired. Because the colors actually printed on the output medium (compared to the colors you are trying to print) vary considerably, and for a variety of known reasons, they drift out of calibration over a period of time (because they drift out of calibration over a period of time (). It will shift). For example, changes in a given or loaded print medium (differences in paper or plastic sheet type, substance, weight, calendaring, coating, moisture, etc.), changes in printer ambient conditions, image development, etc. Aging and wear of materials, printer components, changes in the interaction of different printed colors, etc. Therefore, a print test color patch is highly desirable at the same relative time intervals as when the color print job is color controlled, under the same printing conditions, and on the same print medium test sheet.

【0019】

Therefore, it is also useful to provide a dual mode color test sheet. In a dual-mode color test sheet, multiple color patches of different colors are printed on each or in a given banner, cover, or other inter-document or print job separator sheet, especially in blank areas. Will be done. Different color sets may be printed on different banners or other test sheets. The dual use of such sheets saves both printing paper and printer usage time and also provides frequent color calibration opportunities in printing systems where banner sheets are printed at frequent intervals. To do.

【0020】

An additional feature that may be provided is that a special color or combination of test patches on a special banner or other test sheet is printed on a special document for that banner sheet, i.e., immediately following the banner sheet. Adapting or setting to those colors that are about to be printed on the document page to be printed (the print job identified by its banner sheet). It can provide "real-time" color corrections for color printers that are adapted to correct the color prints of the very next document to be printed.

【0021】

The preferred implementation of the systems and features disclosed herein may vary depending on the circumstances. In addition, various disclosed features or components include features such as grayscale balancing and turning two or more lighting sources, such as facing-to-face LEDs, on at once. Can be used as an alternative.

【0022】

Even if these test patch images and colors are automatically sent to the printer imager from a storage data file specially designed for printing dual-mode banner sheets and other color test sheet pages. It will be appreciated that they may and / or be embedded inside the customer job, including the banner page. That is, the latter may be directly electrically associated with an electronic document to be printed and / or generated or transmitted by the author or sender of the document. Since the colors of the printed test sheet color patches and their printing sequences are known (and stored) information, the online spectrophotometer measurement data from them can be automatically adjusted and compared.

【0023】

After the spectrophotometer or other color sensor has read the color of the test patch, the measured color signal is automatically processed inside the system controller or printer controller, as described in the cited reference material. Create or change a tone reproduction curve. The color test patch on the next test sheet is then printed using the new gradation reproduction curve. This process can be repeated to generate a further modified gradation reproduction curve. The printer's color image printing components and materials are relatively slow, long-term drift, relatively stable, and are generated using this closed-loop control system in the absence of print media or other abrupt changes. The gradation reproduction curve is a correction curve that achieves uniform color for at least one or a considerable number of customer print jobs that are subsequently printed, and is relatively rare and few, such as a normal banner sheet. Only the color test sheet of is required to be printed.

【0024】

In addition to its use in printers, the use of color measurements and / or color measurements for various quality or consistency control functions (of textiles, wallpapers, plastics, paints, inks, foods). It should be noted that it is also important in production, etc., and for many other different techniques and uses (in the measurement and detection of different properties of different substances, objects, or objects). Therefore, the present invention can also be used in various other fields such as these substances, objects, or objects being color tested, where (1) color measurements are performed in a closed loop system. It includes both used and applied uses, and (2) uses where the measurement results are not fed back into the control loop but are used to generate a one-time (single) output.

【0025】

FIG. 1 is a functional block diagram showing an exemplary embodiment of the coloring (color coding, coloring) system 100 according to the present invention. The coloring system 100 is connected to the input device 200 via a link 210. The input device 200 inputs various information required to perform the operation of the coloring system 100, as described in more detail below, and is a mouse, keyboard, touch screen input device, voice recognition. Can include input devices based on, and / or any other known or more recently developed device useful for inputting information. The coloring system 100 is optional, but is connected to the image data source 300 via a link 310. The connection to the image data source 300 is "arbitrary". Because it is only needed for certain embodiments of the coloring system 100.

【0026】

For example, when the coloring system 100 is a marking device such as a printer, an image data source 300 is required. However, when the coloring system 100 is a system for performing a coloring operation that does not require image data, the image data source 300 is not required. An example of a coloring operation that may not require image data is the task of producing colored foods such as grains (colored foods).

【0027】

The image data source 300 is a digital camera, a scanner, a locally or remotely located computer, or any other known or more recently developed device capable of generating electronic image data. May be good. Similarly, the image data source 300 can be any suitable device that stores and / or transmits electronic image data, such as a client or server in a network. The image data source 300 can be integrated with the coloring system 100, as in a digital copier with an integrated scanner. The image data source 300 can also be a modem, a local area network, a wide area network, an intranet, the Internet, any other distributed processing network, or any other known or more recently developed connected device. It can be connected to the coloring system 100 through such a connecting device.

【0028】

It should be understood that electronic image data can also be generated when printing an image from the original physical document, but could have been generated at any time in the past. Furthermore, this electronic image data does not have to be generated from the original physical document, but could also be generated from electronic scratches. The image data source 300 is therefore either a known or more recently developed device capable of supplying electronic image data to the coloring system 100 through the link 310. The link 310 can therefore be either a known or more recently developed device that transmits electronic image data from the image data source 300 to the coloring system 100.

【0029】

Furthermore, it should be understood that links 210, 310 can be wire connections, wireless, optical links to networks (not shown). The network can be a local area network, a wide area network, an intranet, the Internet, or any other distributed processing and storage network.

【0030】

The coloring system 100 includes a coloring device 120, a sensor array 130, a color changing device 140, a memory 150, a controller 160, and a spectrum curve determination system 170 interconnected by a data / control bus 190. The spectrum curve determination system 170 has a reference database 172 and a spectrum curve output device 174.

【0031】

When the coloring system 100 is a printer or other marking device, the coloring device 120 may be, for example, a print engine / printhead or a marking engine / marking head. The coloring device 120 may be, for example, a colorant dispenser that dispenses the colorant onto the object or into the mixture. In a nutshell, the coloring system 120 should be any known or more recently developed device such that an object, substance, or object controls the final appearance directly or indirectly. You can also.

【0032】

The sensor array 130 is arranged around a central photodetector (not shown) or is contiguous with US Application 09 / 862,247, US Application 09 / 863,042, and / or US Application No. It has multiple light emitters such as multiple LEDs and multiple lasers arranged for multiple photodetectors or photosites as described in 09 / 888,791. The light emitter is hereinafter referred to as an LED for convenience. The number of LEDs can be any number greater than 3 if a single photosensor (photodetector) is used, and 2 if multiple photosites or photosensors are used. It only needs to be one. A large number of LEDs provides greater accuracy, but having more LEDs is more costly, and the number of LEDs contained in the sensor array 130 is practically limited. In particular, an object of the present invention is to provide a low cost spectral spectrophotometer. Therefore, the number of LEDs is preferably about 8 to about 16.

【0033】

Each LED is selected to have a narrowband response curve in spectral space. So, for example, 10 LEDs correspond to 10 measurements on the reflectance curve. Equivalents of color sensors based on multiple light emitters, such as LEDs, or lasers, for example, are switched on one at a time, for example, when the measuring medium passes through the transport section of the printer. The reflected light is then detected by a photodetector and the corresponding voltage is integrated and normalized using white tiles. This normalization may be performed periodically. The use of a white tile calibration look-up table stored in memory 150 for normalization is a standard practice in the color measurement industry. When the white tile calibration look-up table is used, the detector output is normalized between 0 and 1 according to the following equation.

【0034】

[Number 2]

<img file="JP2003106900A_D0001.tif" />Where V_i⁰Is the i-th (i^th) LED black measurement detection system offset, V_i^fsIs a white tile measurement, V_iIs the sensor detector output and R_i^wIs the reflection spectrum of the white tile at the average wavelength of the i-th LED. Any other known or more recently developed method for normalization can be used instead. V_mCan be compensated for temperature changes.

【0035】

The color change device 140 calibrates the output of the coloring device 120 according to the information obtained from the spectrum curve output device 174 of the spectrum curve determination system 170. This calibration may be performed to the extent required, i.e., to the extent desired to maintain the desired output of the coloring device 120.

【0036】

The memory 150 can act as a buffer for information entering and exiting the coloring system 100 and contains any program and / or data necessary to perform the functions of the coloring system 100. It is memorable and / or can store data at various processing stages. The white look-up table described above may be stored in memory 150 if desired. The reference database 172 described in more detail below may also be stored in memory 150 if desired. Further, the memory 150 is described as a single entity (component), but may actually be distributed. The variable portion of memory 150 is performed with static or dynamic RAM in various exemplary embodiments. However, the memory 150 can also be executed using a floppy disk and a disk drive, a writable optical disk and a disk drive, a hard drive, a flash memory, or the like. The general static portion of memory 150 is performed with ROM in various exemplary embodiments. However, this static part can be referred to as other non-volatile memory such as PROM, EPROM, EEPROM, optical ROM disk such as CD-ROM or DVD-ROM, disk drive, flash memory, or other as shown above. It can also be executed using a variable memory or the like.

【0037】

The controller 160 controls the operation of the other components of the coloring system 100, performs any necessary calculations, and is required to perform the processing of the coloring system 100 and its individual components. To control the flow of data between other components of the coloring system 100 as needed.

【0038】

The spectrum curve determination system 170 determines and outputs a spectrum curve. In particular, the spectrum curve output device 170 is based on a plurality of spectra determined by the controller 160 based on information from the reference database 172 described in more detail below and the output of the sensor array 130 from a different color target. Output the curve.

【0039】

In order to obtain an output similar to that of a reference spectrophotometer such as the Gretag spectrophotometer, it is necessary to convert the reading from the sensor array 130 into a reflection spectrum. The Gretag spectrophotometer outputs 36 spectral reflectance values. These values are evenly separated at 10 nm over the visible spectrum (eg, 380 nm to 730 nm). Therefore, in the example below, the readings from the sensor array 130 are converted to 36 reflectance values. In other words, when there are 10 LEDs in the sensor array 130, these LEDs are switched in sequence and the readings (generally voltage readings) are collected from the photodetector for each LED. The 10 readings (voltages) from the sensor array 130 are converted to 36 reflectance values for each color. It will be appreciated that when using a multiplex photosite sensor, the desired number of outputs, eg 10 outputs, can be obtained from a smaller number of LEDs, eg 3 or 4 LEDs.

【0040】

The reference database 172 measures the reflection spectra of several reference color sets using an accurate reference spectrophotometer, such as the Gretag spectrophotometer, and their corresponding LED sensor outputs using the sensor array 130. It is caused by that. In general, the more databases you have, the more reference colors you use, the more accurate the results will be. In one exemplary reference database, about 2000 colors are used. Measurements of the reference spectra of several reference color sets and their corresponding LED sensor outputs generate database 172. The data stored in the reference database 172 is hereinafter referred to as a training sample. The reference database 172 may be compressed and only the cluster centers described below (center of the population) need to be stored inside the sensor controller hardware. Details on how the reference database 172 is used are given below.

【0041】

It will be appreciated that each circuit shown in Figure 1 can be implemented as part of a well-programmed general purpose computer. In addition, each circuit shown in FIG. 1 uses a separate logic element or a separate circuit element as a physically separate hardware circuit inside the ASIC, or using FPGA, PDL, PLA, or PAL. Can be implemented. The special form that the individual circuits shown in Figure 1 will take is design freedom, which will be obvious and predictable to those skilled in the art.

【0042】

An exemplary algorithm that can be implemented by controller 160 to determine the spectrum based on the output of reference database 172 and sensor array 130 is described below. The following algorithm is a fuzzy inference-based spectrum reconstruction algorithm. This fuzzy inference-based algorithm determines the spectrum of a given color sample from a non-linear model, such as a fuzzy model, calculated (computer) from a training sample database (reference database 172). Basically, a non-linear model, such as a fuzzy model, calculates (normalizes) each point of the reconstructed spectrum (by computer) from the weighted sum of the normalized voltages generated from the color samples. The calculated voltage is calculated by Eq. (1)). Each of these total weights is a function of the corresponding normalized voltage. Therefore, these weights are constant and therefore the algorithm is "non-linear".

【0043】

In the following description, the number of LEDs contained in the sensor array 130 is assumed to be 10. Those skilled in the art will be able to understand how the algorithm is applied to sensor arrays with more or fewer LEDs.

【0044】

Further, from such a context, it should be understood that, in general, an algorithm that can be applied to the generation of a gradation reproduction curve cannot be applied to the generation of a spectral curve. One reason is that the first and last values of the gradation reproduction curve are known (ie they are [0,0] and [255,255]), but the spectral curve generated using the LED sensor. The same cannot be said about. This is because the LEDs at both ends of the spectrum (ie, the blue and red LEDs) are not monochromatic.

【0045】

The fuzzy inference algorithm is a complex multidimensional algorithm when applied to the multiple LED sensor array 130 described above. As an example of the algorithm, a simple one-dimensional version of the algorithm is described below for ease of understanding. The basic principle of a one-dimensional algorithm is explained by modeling the output reflectance at an average LED wavelength, as measured by illuminating a single LED.

【0046】

Due to the wideband characteristics of LEDs, it is difficult to accurately obtain true reflectance values at average LED wavelengths by performing unit calibration on white tiles. If a detection system model is formed at the average LED wavelength and then a modified lookup table is formed using that model, the output of the detector for one LED illumination (LED illumination) at that wavelength. It can be assumed to emit a nearly true reflectance. This modified lookup table is a 1 input 1 output table when there is only a single LED. How to build such a one-input, one-output lookup table is shown below.

【0047】

Figure 2 shows the normalized output voltage of an LED sensor for various color samples as a function of the corresponding reference spectrophotometer output when only one blue LED is illuminated. The normalized LED voltages are on the x-axis and their corresponding reflectances are on the y-axis as measured by a reference spectrophotometer such as the Gretag spectrophotometer. Data point 500 represents experimental data obtained individually from different experiments.

【0048】

From these experimental data, cluster centers 510, 520, 530, 540 are determined. These cluster centers 510, 520, 530, 540 are shown as large circles, and each circle actually represents a single point on the graph that corresponds to the center of the circle. The actual number of cluster centers can be determined by the user, or can be determined automatically as described in more detail below. These centers, when used properly, can also represent the original data points, as described below.

【0049】

Each cluster center 510 to 540 has related clusters, and any point 500 in FIG. 2 belongs to one or more clusters. The main idea of clustering is to represent any given point in Figure 2 as a combination of cluster centers or a weighted sum. As a given point approaches the center, the weighting of this center becomes heavier. For example, considering the four cluster centers 510-540 shown in Figure 1, any point in the dataset (x = 0.10, y = 0.12) belongs only to the cluster associated with the cluster center 510. It can be assumed (because this point basically coincides with this cluster center) and is therefore represented only by the cluster center 510. Similarly, points (x = 0.22, y = 0.20) belong to both the cluster associated with cluster center 510 and the cluster associated with cluster center 520 and are represented by the combination of cluster center 510 and cluster center 520. (Because this point is located between these two cluster centers).

【0050】

Either known or more recently developed algorithms can be used to obtain a cluster center. For example, to obtain a cluster center, the algorithm described in the article by Stephen L. Chiu, "Fuzzy Model Discrimination Based on Cluster Estimates", which is incorporated herein by reference in its entirety, can also be used. ..

【0051】

After the cluster centers 510 to 540 are obtained, the spectral reflectance values using one LED illumination can be reconstructed using the cluster centers 510 to 540 and the logic for assigning weights. The model can be reconstructed by a weighted sum of cluster centers 510-540 for a given input voltage from a single LED light. Weighting is assigned to each of the cluster centers based on the assigned (or perceived) "influence" of the input LED voltage, which is the function shown at the top of Figure 2. Performed by using 600. Of these functions 600, the functions represented by lines 610 and 620 represent influences on the scale 0 to 1 of cluster center 510, and the functions represented by lines 650 and 660 represent influences of cluster center 530. And the functions represented by lines 670 and 680 represent the influence of the cluster center 540.

【0052】

The model has an estimated reflectance R for any given input from one LED light._estIs output. For example, as shown in FIG. 2, of the 500 data points, the point with a normalized LED voltage of 0.15 is aligned with the point 625 on line 620. At point 625, the value of line 620 is 0.8. Therefore, the cluster center 510 has an influence of 0.8 at an LED voltage of 0.15, which is aligned with the point 635 on the line 630. At point 635, the value of line 630 is 0.2. By the way, the cluster center 520 has a 0.2 influence on the data point 500 at an LED voltage of 0.15. Correspondingly, the estimated reflectance value R for the normalized LED voltage_estIs R_est= 0.8 * R₁+ 0.2R₂Can be defined by. Where R₁Is the y-component (reflectance) of cluster 510, R₂Is the y-component (reflectance) of the cluster center 520. This expression corresponds to a 1-input, 1-output lookup table.

【0053】

Function 600 has a single shape in this example. In general, the functions actually used in this algorithm have a more complex shape.

【0054】

The model described above can be regarded as a fuzzy inference system (FIS). More information about FIS can be found in any book on fuzzy logic, such as The Math Works Inc. Fuzzy Logic Toolbox for use with Matlab--User's Guide Version 2 (Natick MA, January 1999). it can.

【0055】

As described above, a number of cluster centers, such as the cluster centers 510-540 described above, and their corresponding membership functions are used to model a given sensor. This model can be thought of as a large number of if-then rules, where the number of rules is equal to the number of cluster centers. For the four cluster centers, the corresponding if-then rules are: [Number 3]

<img file="JP2003106900A_D0002.tif" />Is. Where R is the estimated spectral reflectance and V_mIs the normalized LED sensor output, R₁, R₂, R₃, R₄Is the spectral reflectance value for the cluster center, as measured by a reference spectrophotometer, μ₁, Μ₂, Μ₃, Μ₄Are the membership functions of the 1st, 2nd, 3rd, and 4th cluster centers, respectively. V when 10 LEDs are used_mIs a 10-component vector, 8-component vector when 8-component is used, and so on.

【0056】

When using the if-then rule described above, the estimated reflectance value R for a given LED sensor input Z is: [Number 4]

<img file="JP2003106900A_D0003.tif" />Can be obtained by

【0057】

Therefore, Eq. (3) gives the estimated reflectance value R for a single LED illumination. For example, when only two functions affect a given sensor input Z, as in the example above in FIG. 2, it will be understood that the values of the other two functions in equation (3) are 0.

【0058】

When a single LED is used, the reconstructed reflectance value R can be obtained simply by using Eq. (3) each time a new measurement is needed. Membership function μ₁, Μ₂, Μ₃, Μ₄Is stored inside the memory 150 (Fig. 1).

【0059】

The fuzzy inference algorithm applied in the case of a single LED sensor is described above to facilitate the understanding of the fuzzy inference system applied to general LED sensors. Next, the fuzzy reasoning principle described above is extended to a multiple LED spectrophotometer in order to reconstruct all spectra using all readings from a switchable multiplex LED device such as the LED sensor array 130 described above.

【0060】

The input of the algorithm is an n-dimensional vector with n normalized voltages from the LED sensor array 130 (see Equation (1)). The output of this algorithm is an l (L) dimensional vector with estimates for the reflection spectra at l different wavelengths. The example described below uses 10 LEDs, and since this reference spectrophotometer is assumed to be a Gretag spectrophotometer with 36 outputs, it goes from 10 to 36 (10-to). -36) Reconstruction is shown. Therefore, in this example, n = 10 and l = 36. However, this method is commonplace and can be applied to any number of inputs n and outputs l.

【0061】

To implement a reconstruction algorithm based on fuzzy inference, a cluster center is first obtained. The number of cluster centers can be arbitrarily determined by the user, and the coordinates of the cluster centers may be entered manually, for example by inspection of experimental data. Further, the number of cluster centers and the coordinates of the cluster centers may be automatically obtained. FIG. 3 shows an example of a method for automatically determining the number of cluster centers and the coordinates of the cluster centers. The algorithm shown in FIG. 3 is the "subtractive clustering algorithm". Further details on the principles of this algorithm can be found in the above-mentioned paper by Chiu.

【0062】

In FIG. 3, the algorithm starts at step S1000 and continues to step S1100. N data points x in step S1100₁, x₂, .... x_NCollection is obtained. Each of these data points represents a mapping from the LED output voltage to the reference spectrophotometer reading, such as data point 500 in FIG. Acceptance ratio E, which can be entered or automatically determined by the user_lAnd weighting factor r_aAnd σ are also obtained. User selection or automatic determination may be based on empirical data, or may be based on one or more predetermined criteria. Acceptance Ratio E_lAn example of a suitable value for is 0.5. An example of a suitable value for the weighting factor σ is 1.25. Weight coefficient r_aAn example of a suitable value for is 0.95. Of these, the weighting factor r_aIs the most important. Because it affects the required if-then rules and also affects the prediction error. r_aA suitable value for is r_aExperimental value of r_aGraph the corresponding expected error values for each value of r_aGraph the experimental values of for the corresponding number of required rules, (1) the best in terms of both the number of required rules and the expected error, or (2) required R that is one of the appropriate compromises between the number of rules and the expected error_aIt can be determined by selecting the value of.

【0063】

Each data point x_iIs a vector with M components, where M = n + l (= 10 + 36 = 46 (in the case of 10-to-36 reconstruction)). Each data point x_iIs assumed to be normalized in each dimension (size, dimension).

【0064】

Algorithm parameter r_a, Σ, E_lIs selected and the data point x₁~ x_NIs obtained in step S1100, then the algorithm proceeds to step 1200, where the potential P for each of the N data points._i(On a computer), [Number 5]

<img file="JP2003106900A_D0004.tif" />Calculated by. According to equation (4), a data point having many adjacent (proximity) points has a high potential, while an isolated data point has a low potential. Parameter r_aIs the weighting factor. r_aIf is larger, the "influence radius" or each x_iThe neighborhood (adjacent part) of is also larger.

【0065】

Then, in step 1300, the algorithm has the highest potential P of the N data points._kData point with x_kSelect. This data point is the first (first) cluster center C_x1Will be. The algorithm is P_k= P_c1Set to, and then step S1400 continues. At step 1400, point x_kIs removed from the cluster center candidates. The algorithm then continues to step S1500.

【0066】

At step 1500, the potential of the remaining data points is [Number 6]

<img file="JP2003106900A_D0005.tif" />Is calculated again (on the computer) by.

【0067】

Then, in step S1600, the highest potential P_kData point with x_kIs the next cluster center C_x2Is selected as a candidate as much as possible. X in step S1600_kAnd P_kIs x in step S1300_kAnd P_kIt will be understood that it is different from. Because the original data point x selected in step 1300_kIs a cluster center and potential P_kIs removed from the candidate pool after being recalculated. Also, Cluster Center x_kIt can also be seen that the potential for data points close to is significantly reduced. The parameter σ is a weighting factor for this reduction. The algorithm then continues to step S1700.

【0068】

In step 1700, the data points selected in step 1600 x_kIs determined to be eligible as a cluster center. This judgment is made on the basis of one or more predetermined criteria. As a simple example, the criteria for step S1700 is P_k&gt; E_l* P_c1Is. However, more advanced (complex) criteria may be applied, such as those described in the paper by Chui mentioned above.

【0069】

If the criteria are met in step S1700, the algorithm follows step S1800, x_kAs a cluster center, return to step S1400 and repeat steps S1400 to S1700. If the criteria are not met, the algorithm proceeds to step S1900 and exits. The number of cluster centers determined at the stop point in step 1900 is the number of cluster centers used in the fuzzy inference algorithm described below._cIs.

【0070】

Here we assume that we have N sets of color measurements in a reference database 172 that represents color measurements from LED-based devices and reference spectrophotometers. For example, if 1000 Pantone® colors are used as samples to model the sensor, N represents 1000 LED sensor outputs-reflectance data set. In each dataset, the ED output contains 10 values and the reflectance contains 36 values for a 10-LED sensor, a Gretag spectrophotometer or another reference spectrophotometer with 36 outputs. As mentioned above, each database x_iIs represented as a vector of M components. Where M = n + l, the first n = 10 components are the LED readings, and the last l = 36 components are the reference spectrophotometer (ie, Gretag) readings. In other words, about the reference database x_i= [y_i, Z_i]; i = 1, ...., N (6) And y_iIs an n-dimensional input to the system to be modeled, z_iIs the l-dimensional output from this system. Figure 4 shows a simple scheme for this. Similarly, n_cCluster Center C_xjEach [0071]

[Number 7]

<img file="JP2003106900A_D0006.tif" />It is divided as shown by. Where C_yjRepresents 10 components for each 10-LED sensor, C_zjEach represents 36 components for one 36-power reference spectrophotometer.

【0072】

The cluster center is used to model the system and makes it possible to predict the system output for a given input. Intuitively, Cluster Center C_xjInput part C_yj"Closed" input to V_mIs the output part C of this cluster center_zjProduces a "close (closed)" system output to. This idea is detailed below.

【0073】

Division C as defined in equation (7)_yj, C_zjConsider a set of centers with. V any system input vector_mAssuming V_mIs n_cThe degree to which each of the clusters belongs [Number 8]

<img file="JP2003106900A_D0007.tif" />Shall be given by. Here μ_j(V_m) Is a membership function used in multidimensional algorithms. The membership function given in Eq. (8) is just one example of a suitable function. See one of the books on fuzzy systems for other examples of eligible membership functions.

【0074】

Input V_mEstimated system response to (estimated spectrum (estimated spectrum)) [Number 9]

<img file="JP2003106900A_D0008.tif" />Is modeled by the following weighted means.

【0075】

[Number 10]

<img file="JP2003106900A_D0009.tif" />here, [Number 11]

<img file="JP2003106900A_D0010.tif" />And μ_j(V_m) Is defined in equation (8), C_ZjIs defined in equation (7).

【0076】

[Number 12]

<img file="JP2003106900A_D0011.tif" />Other solutions to are described below and also described in the above-mentioned paper by Chiu. In this solution, C as in equation (9)_ZjInstead of using a linear model [0077]

[Number 13]

<img file="JP2003106900A_D0012.tif" />V defined by_ZjTo use. Here G_jIs a constant matrix, h_iIs a constant vector of compatible dimensions. These G_jAnd h_jIs calculated by solving the least squares estimation problem, as outlined below. C in equation (9)_ZjV instead of_ZjWith [0078]

[Number 14]

<img file="JP2003106900A_D0013.tif" />Will be. All data points (y_j, z_i) Evaluate this formula, G_j, H_j,j = 1, ...., n_cGiven the N equations in. These N formulas X = Y * A ... (13) It can be written in a compact form. Where X is the output z_j Is an N × l matrix calculated by replacing, where Y is ρ_jAnd y_iCalculated from N × (n_c+ n * n_c) It is a matrix. A is an unknown G_jAnd h_jIncluding (n_c+ n * n_c) × l matrix.

【0079】

In general, A cannot be like X = YA, so the least squares solution A_LsIs A_LS= (Y^T* Y)^-1* Y^T* X ... (14) Calculated by (assuming Y is a full-column rank matrix). Where Y^TIs the transpose of the matrix Y.

【0080】

In summary, the LED normalized voltage (or vector) is V_mAssuming that the reconstructed spectrum (or output vector) [Number 15]

<img file="JP2003106900A_D0014.tif" />The points of are calculated using Eq. (12). Where ρ_jIs given by equation (10), G_jAnd h_j(Obviously) A in Eq. (14), using X, Y calculated from the training samples contained in the reference database 172._LSDerived from.

【0081】

FIG. 5 is a flowchart summarizing the steps of the first exemplary reconstruction algorithm according to the invention. Starting at step S2000, this process continues to step S2100. Here, the training sample is input from the reference database, and an appropriate value is r._aEntered about, n_cCluster Center C_xjIs obtained as described above. This process then follows step S2200, where the cluster centers are partitioned as described above, C._yj, And C_zjIs obtained. This process then proceeds to step S2300 to receive sensor readings, such as sensor voltage, from each light emitter in the sensor array. This process normalizes the sensor readings obtained from the sensor array based on the white tile calibration look-up table._mTo get.

【0082】

Then, in step S2400, the process follows equation (8) above to V for each cluster center._mMembership μ_j(V_m) Is determined. This process then follows step S2500, weighted ρ according to equation (10) above._j(V_m) Is determined. The process then continues to step S2600. Where the spectrum [Number 16]

<img file="JP2003106900A_D0015.tif" />Is weighted ρ using, for example, Eq. (9) above._j(V_m) And division C_zjIt is decided based on. Following step S2700, it is determined whether all color samples have been measured. If not all color samples have been measured, the process continues with step S2800. Otherwise, the process jumps to step S2900.

【0083】

In step S2800, the next color sample is selected. Steps S2300 ~ S2700 are then repeated. When all the color samples have been measured, the process proceeds to step S2900 and outputs the full reflectance spectra of the color samples, that is, the spectral curves. Finally, the process ends at step S2990.

【0084】

FIG. 6 is a flowchart summarizing the steps of the second exemplary reconstruction algorithm according to the invention. Steps S3000 to S3990 in FIG. 6 are the same as steps S2000 to S2990 in FIG. 5, with the following exceptions. In step 3100, enter the training sample from the reference database and r_aGet a reasonable value for n_cCluster Center C_xjIn addition to obtaining, this process additionally involves the parameter G, as described above in relation to equations (13) and (14)._jAnd h_jTo get. In step S3600, the spectrum [Number 17]

<img file="JP2003106900A_D0016.tif" />Is weighted ρ using Eq. (12) above._j(V_m) And parameter G_jAnd h_jIt is decided based on.

【0085】

The nonlinear modeling technique described above can be regarded as a fuzzy inference system (FIS) in which each cluster center gives fuzzy if-then rules. Therefore, using Eq. (9), FIS has the following form of n:_cHave rules.

【0086】

IF Y1 is Q_in1 AND Y2 is Q_in2 AND ... AND Y_NIs Q_inn, THEN Z is Q_outj AND Z₂Is Q_outj2... ANDZ_LIs Q_outjlWhere Y_kIs the input vector V_mKth component of (V_mk), Z_rIs the output vector [Number 18]

<img file="JP2003106900A_D0017.tif" />Represents the r-th component of_injk(For the jth rule) [Number 19]

<img file="JP2003106900A_D0018.tif" />Is a membership function like, Q_outjrIs Q_outjr(Z_r) = C_zjrGiven by.

【0087】

The AND operator is multiplication and the rule output is weighted to calculate the final output.

【0088】

FIG. 7 is a functional block diagram showing an exemplary embodiment of the color detection system 500 according to the present invention. The color detection system 500 has an input / output interface 110, a sensor array 130, a controller 150, a memory 160, and a reference database 172, which may be identical to the corresponding elements of FIG. 1 incorporated by the data / control bus 590. The color detection system 500 is connected to the user input device 200 via the link 210 in the same manner as the input device 200 and the link 210 described above with reference to FIG. The color detection system 500, like the links 210 and 310, is also connected to the data sink 400 via a link 410, which can be a wire connection, wireless, or optical link to the network. In general, the data sink 400 can output or store processed data generated by a color detection system, such as a printer, copier, or other forming device, facsimile device, display device, memory, etc. It can also be a device.

【0089】

The color detection system 500 may or may be included in a portable or fixed unit designed specifically for measuring the color of the target object. In use, the color detection system 500 is positioned with the sensor array 130 facing the target object, the sensor array 130 operates as described above, and then the algorithm described above is an estimation spectrum of the target object. [Number 20]

<img file="JP2003106900A_D0019.tif" />To obtain, run by controller 150 with data from sensor array 130 and reference database 172. Estimated spectrum [Number 21]

<img file="JP2003106900A_D0020.tif" />Is then output to the data sync 400.

【0090】

From the above description, in these embodiments, the invention is capable of providing calibration tools for scanners, printers, digital photocopiers, etc. Also, in these embodiments, the invention is: It will be appreciated that it is possible to provide a color measurement tool designed to provide one-time color measurement of multiple target objects.

【0091】

The coloring system 100 of FIG. 1 and the color detection system 500 of FIG. 4 are programmed using the associated sensor array 130 (and the coloring device 120 in the case of FIG. 1) on a single general purpose computer or separately. It is preferably carried out on any of the above-mentioned general-purpose computers. However, the coloring system 100 and the color detection system 500 are hardware such as dedicated computers, programmed microprocessors or microcontrollers and peripheral integrated circuit elements, ASICs or other integrated circuits, digital signal processors, and separate element circuits. It can also be implemented on wired electronics or logic circuits, programmable logic devices such as PLDs, PLAs, FPGAs, PALs, etc. In general, a device capable of carrying out a finite state machine capable of carrying out a flowchart as shown in FIGS. Can be used to implement a build device.

【0092】

In addition, the disclosed methods are readily implemented in objects that provide portable source code that can be used on a variety of computer and workstation hardware platforms, or in object-oriented software development environments. obtain. Also, suitable parts of the disclosed coloring system 100 and color detection system 500 can be partially or completely implemented in hardware using standard logic circuits or VLSI designs. Whether the software or hardware is used to implement the system in accordance with the present invention depends on the speed and / or efficiency requirements of the system, special features, and the special hardware or hardware being used. Depends on the system or microprocessor or microcomputer system. The processing systems and methods described above, in hardware and software, use either known or more recently developed systems or configurations, devices and / or software by those skilled in the art, along with general knowledge of computer technology. From the description of the functions described herein, it can be easily carried out without performing more experiments than necessary.

【0093】

Furthermore, the disclosed method can be easily implemented as software execution on a programmed general purpose computer, dedicated computer, microprocessor or the like. In this case, the methods and systems of the invention are resources that exist as routines embedded on personal computers, such as photocopiers, color photocopiers, printer drivers, scanners, etc., or on servers or workstations. Can be implemented as a (resource). These systems and methods can also be implemented by software and / or physical integration into the hardware system, such as hardware and software systems for photocopiers or dedicated image processing systems.

[Simple explanation of drawings]

[Figure 1]

It is a functional block diagram which shows the exemplary embodiment of the coloring system by this invention.

[Figure 2]

It is a figure which shows the normalized output voltage of a single LED sensor as a function corresponding to the reference spectrophotometer output.

[Fig. 3]

It is a flowchart which shows the exemplary method of obtaining a cluster center.

[Fig. 4]

It is a figure which outlines an exemplary process which reconstructs a spectrum from the reading of an LED sensor.

[Fig. 5]

FIG. 5 is a flow chart illustrating a first exemplary method of determining a spectrum according to the present invention.

[Fig. 6]

It is a flowchart which shows the 2nd exemplary method of determining a spectrum by this invention.

[Fig. 7]

It is a functional block diagram which shows the exemplary embodiment of the color detection system by this invention.