JP5608511B2 - Image correction apparatus and image correction method - Google Patents

Description

  The present invention relates to a technique for removing a shadow of an image captured by a camera or a scanner.

  2. Description of the Related Art Conventionally, in offices and homes, there are techniques for capturing documents, slips, postal items, banknotes, and the like with a camera or scanner using an area image sensor and storing or using the obtained image data.

  Also, for example, in a financial institution, a scanner connected to a terminal device installed at a sales office window captures an image of a transaction form filled by a customer and stores the obtained image data, Character recognition is performed. In that case, a stand-type scanner is often used as the scanner.

  The stand-type scanner is a non-contact type scanner that reads a document with a reading surface facing upward from an image sensor fixed to a support from above. Compared with flatbed scanners, stand-type scanners can be used more often in the counter operations of financial institutions that handle a variety of forms because they can reduce the time and effort required to scan paper of various sizes and thicknesses. ing.

  However, when a stand-type scanner is used, shadows (brightness unevenness) may occur in the read image data depending on illumination conditions and the like. Therefore, it is necessary to remove the influence of ambient light by performing calibration. As a general calibration method, shading data representing shadow density is acquired by placing a blank sheet on a platen and reading it, and shading correction of the document image is performed using the shading data. As another calibration method, for example, there is a method of performing calibration based on an image on a platen instead of a blank sheet (see Patent Document 1).

  However, there may be a case where a shadow is generated in the read image due to a change in the illumination environment after the calibration is performed or a change in the surrounding situation such as a person standing near the scanner during imaging. Such a shadow is hereinafter referred to as a “dynamic shadow”. Dynamic shadows have a problem of adversely affecting image processing, such as a decrease in accuracy of character recognition processing using captured images.

  Patent Document 2 discloses a technique for removing dynamic shadows from a scanner image or a camera image. In this technique, an image is divided into partial areas, and shadow determination and density determination are performed for each partial area. Specifically, the frequency distribution of the luminance values of the pixels in each partial area is taken, and the peak luminance close to the white level is detected as the representative value of the luminance of the document ground color (ground color including the influence of shadows). To do. This value is the luminance value of the ground color at the center coordinates of the partial area.

  The luminance value of the ground color at other coordinates can be obtained from the luminance value of the central coordinates of the partial region using linear interpolation. It can be determined that the lower the luminance value of the ground color, the darker the shadow. In the luminance value correction, a correction value for each pixel is obtained based on a ratio between a predetermined luminance value serving as a white reference (for example, 255 indicating a white pixel) and the ground color luminance, and is corrected to the luminance of the pixel of the input image. It can be obtained by multiplying the values.

JP 2008-199556 A JP 2002-15313 A

  In general calibration, correction can be performed using blank paper as a reference, but there is no image serving as a correction reference for correcting dynamic shadows. Therefore, it is necessary to detect and correct the shadow area and density only with the captured image. As described above, in the technique of Patent Document 2, in order to remove a shadow without a reference image, a peak close to white is used as the luminance of the ground color from the frequency value luminance distribution.

  However, since this method is based on the premise that the background color of the document is uniform white, there is a problem that it cannot be applied to a color document. For example, consider a case in which a red or green ground color region is printed on white paper and the document is applied to a document having a small white area. In this case, since the luminance value of the ground color is different for each region, the peak corresponding to the red or green ground color is detected in the frequency distribution of the luminance value, but it may not be detected as a peak because the area of the white paper is small. is there. In such a case, it is not possible to appropriately detect paper white as the background color. Furthermore, when the luminance is corrected using a printing color different from the paper color as a ground color, there is a problem that this color region also becomes white. In order to solve these problems, it is necessary to determine the ground color and correct the luminance in the color document.

  Accordingly, the present invention has been made to solve such a problem, and an object thereof is to appropriately remove dynamic shadows in a color image obtained by imaging a subject.

In order to solve the above-described problems, the present invention provides an image correction apparatus that performs correction for removing a shadow from a color image obtained by imaging a subject with a non-contact scanner, the color image, and a plurality of images A storage unit that stores a predetermined determination criterion for detecting the background color of the subject from the representative color, and after dividing the color image into a plurality of partial areas, each pixel in each of the partial areas is represented in a color space. It performs color clustering is a process of clustering determined representative color of each cluster using the results of the color clustering by the predetermined criteria to detect any of the determined representative color as a background color, on the basis of the detected partial regions each of the background color, to produce a ground-color-density map showing a ground color density of the entire image region corresponding to the object, use the ground color density map An image correction unit that performs correction to remove a shadow from each pixel value in the color image, assuming that the darker the background color is, and stores the corrected pixel value in the storage unit. It is characterized by.

  ADVANTAGE OF THE INVENTION According to this invention, the dynamic shadow in the color image obtained by imaging a to-be-photographed object can be removed appropriately.

It is a block diagram of the image correction apparatus of 1st Embodiment. It is the schematic of the stand type scanner which is an example of an imaging part. It is a flowchart which shows an image correction process. It is a figure which shows the example of the color form image used as correction | amendment object. It is a figure which shows the example of the color clustering of a partial area. It is a figure which shows the example of the frequency distribution of the brightness | luminance in the partial area | region where there are few areas of the ground color of paper. It is a figure which shows the example of the color clustering of the partial area | region where the area of the ground color of paper is small. It is a block diagram of the image processing apparatus which performs image correction and character recognition. It is a block diagram of a sales office window terminal. It is a figure which shows the example of a ground color density map (before interpolation). It is a figure which shows the example of a ground color density map (after interpolation).

  DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, modes for carrying out the present invention (hereinafter, referred to as embodiments) will be described by dividing them into first to sixth embodiments with reference to the drawings (refer to drawings other than the reference drawings as appropriate). To do.

(First embodiment)
First, the configuration of the image correction apparatus according to the first embodiment will be described. As illustrated in FIG. 1, the image correction apparatus 100 corrects a shadowed color image acquired by capturing a form, a document, or the like to correct a shadowless color image (an image equivalent to an image captured in a situation without a shadow). And includes an operation unit 110, a display unit 120, an imaging unit 130, a storage unit 140, and a control unit 150.

  The operation unit 110 is an operation unit including a keyboard used when a user performs input operations of numbers and characters, a button for instructing start of scanning, and the like. The display unit 120 is a display unit including a liquid crystal display that displays an input image, a corrected image, and the like. The imaging unit 130 is an imaging unit such as a digital camera or a stand-type scanner (non-contact scanner) that captures an image using an imaging element such as a CCD (Charge Coupled Device).

  The storage unit 140 is a storage device including a memory, a hard disk device, and the like, and stores images captured by the imaging unit 130, various internal processing programs, and various setting data. Further, for example, as a predetermined determination criterion for detecting the ground color of a subject (such as a document) from a plurality of representative colors, the storage unit 140 selects a color having the highest luminance among the obtained representative colors of each cluster. Memorize the standard of ground color.

The control unit 150 is a control unit that performs image capture processing and input image correction processing, and includes an image capture processing unit 151 and an image correction unit 152.
The imaging processing unit 151 is a control unit that performs imaging control of a subject using the imaging unit 130. When the subject is imaged by the imaging unit 130, the imaging processing unit 151 uses a calibration parameter such as shading data other than dynamic shadows. The image data from which the luminance unevenness is removed is developed as an input image on a memory and temporarily held, and the image data can be stored in the storage unit 140 as a file. As an example of a processing method of the imaging processing unit 151, the method described in Patent Document 1 described above can be cited.

  The image correction unit 152 is means for removing a dynamic shadow from the input image created by the imaging processing unit 151, and is a characteristic part in the image correction apparatus 100 of the first embodiment. Specifically, the image correction unit 152 divides the image into a plurality of regions, performs color clustering (details will be described later) in each region, and selects a color having the maximum luminance among the representative colors of the plurality of clusters. The area color. Further, the luminance value of the pixel is corrected using the luminance (density) of the ground color. Specific processing will be described later with reference to FIGS.

  Next, an example of the imaging unit 130 will be described. As shown in FIG. 2, a stand-type scanner 200 that is an example of the imaging unit 130 includes a camera head 210, a column 220 that fixes the camera head 210, and a document table 230 on which a document 240 that is a subject is placed. The imaging direction of the camera head 210 is downward, and the document 240 placed on the document table 230 can be read. In addition, although the stand-type scanner 200 was mentioned as an example of the imaging part 130, you may use a normal digital camera, a mobile phone camera, etc. in addition to this.

  Next, the processing procedure of the image correction unit 152 will be described in detail. It is assumed that image information obtained by correcting the image collected by the imaging unit 130 using the calibration parameters by the imaging processing unit 151 is an input image, and the input image is stored in the storage unit 140 in advance. This input image may have a dynamic shadow due to illumination fluctuation after calibration or the presence of a person near the imaging unit 130. The processing flow of FIG. 3 aims to remove such dynamic shadows. The contents of the processing in each step are as follows.

  As shown in FIG. 3, the image correction unit 152 first divides the entire image into a plurality of partial regions (step 300). In general, it is divided into rectangular areas of N × M pixels. Specifically, for the convenience of processing accuracy and speed, for example, one partial region may be divided so as to have a size of about 200 × 200 pixels, but may be divided into other sizes.

  Thereafter, the image correction unit 152 executes the processing of color clustering (step 320) and ground color detection (step 330) for all partial areas (“No in step 310 → step 320 → step 330”). repetition). Note that division into partial regions is not essential, and the subsequent processing may be performed on the entire input image without performing region division (step 300). For example, when the image itself is originally small, or when there is a high probability that the shadow formed in the entire image is uniform, the region division may not be performed.

  The image correction unit 152 performs color clustering for each partial region (step 320). Color clustering is a process of clustering each pixel in a processing area on a color space. More specifically, after mapping each pixel on the color space, it is possible to divide the color space into regions using the Voronoi division, Euclidean distance threshold, or the like as the same clusters as the same cluster. In each cluster, the color located at the center of the cluster or the color with the highest frequency can be set as the representative color of the cluster.

  Here, the color clustering will be specifically described using the form image example of FIG. 4 and the color clustering example of FIG. A color form 400 shown in FIG. 4 is printed with a ground color area 410 in which the ground color is the same as white, which is the original ground color of paper, a ground color area 420 in which the ground color is printed in red, and a ground color in green. The background color area 430 is included, and each area includes a white area in which characters indicated by reference numeral 440 and the like are written.

  FIG. 5 is an example in which the colors of the pixels in the partial region 450 of FIG. 4 are mapped to the RGB (Red, Green, Blue) color space. The RGB color space is a three-dimensional space having the values of RGB as axes, and the values of the axes of RGB range from 0 to 255. The origin (0, 0, 0) represents black and (255, 255, 255) represents white. Each pixel in the processing target area can be mapped to one point on the RGB color space. As a result, a plurality of distributions centered on representative colors can be formed, and each can be regarded as a cluster.

  When the partial area 450 in FIG. 4 and each cluster in FIG. 5 are associated with each other, the cluster 500 is a white distribution that is the paper ground color of the color form 400, the cluster 510 is a black distribution that is the character color, and the cluster 520 is the ground color. The red distribution in the color area 420 and the cluster 530 represent the green distribution in the ground color area 430.

  In FIG. 5, the RGB color space is used, but instead, the L * a * b * color space, the L * u * v * color space, HSV (Hue, Saturation, Lightness ( Value)) color space, HSL (Hue, Saturation, Lightness) color space, etc. may be used.

  Furthermore, color clustering may use all pixels in a region or some pixels. For example, pixels sampled at regular intervals may be used for color clustering. Furthermore, in this example, pixel values are directly subject to clustering, but in order to reduce the influence of large color fluctuations with neighboring pixels, such as false colors that occur due to CCD characteristics, An average value or a weighted average value may be obtained and used for color clustering.

The ground color detection (step 330) is a process for detecting the ground color of the partial region using the color clustering result. In the first embodiment, the color having the highest luminance among the representative colors of the clusters obtained by color clustering is set as the ground color. In order to calculate the luminance Y from the pixel color (RGB value) in the RGB color space, for example, the following equation (1) may be used.
Y = 0.299 × R + 0.587 × G + 0.114 × B (1)

  However, this formula (1) is an example, and other formulas may be used. Such a conversion formula is for calculating the brightness, which is the degree of brightness that human vision perceives from the color of the pixel in the RGB color space, and there is no reason limited to this formula (1). It is.

  Here, the reason why the color having the highest luminance among the representative colors of the clusters obtained by color clustering in the first embodiment is used as the ground color will be described with reference to FIGS. When the ground color is determined from the partial region 460 of FIG. 4, if the frequency distribution of luminance values is taken using the technique of Patent Document 2, the result is as shown in FIG. In FIG. 6, a peak 600 represents a character, a peak 610 represents a red ground color region 420, a peak 620 represents a green ground color region 430, and a protrusion 630 represents a paper ground color. Since the area of the white ground color has a small area, no peak appears, and in the method of setting the peak value of the maximum brightness to the ground color density, the brightness value 640, which is a representative value of the green ground color density, is used as the ground color. I will choose. In contrast, FIG. 7 shows the result of color clustering of the partial region 460 of FIG.

  The cluster 700 is a distribution of white (white in the region of the number “8”) that is the ground color of the paper of the color form 400, and the cluster 710 is a character color (here, a frame of the ground color region 420 equivalent to the character color) The distribution of black as a line color), the cluster 720 represents the red distribution of the ground color region 420, the cluster 730 represents the green distribution of the ground color region 430, and the cluster 700 as a white ground color region and the green distribution The cluster 730 that is the ground color region is separated. In the method of using the maximum luminance of the representative color of the cluster according to the first embodiment as the ground color density, it can be determined that the luminance value 650 (see FIG. 6) of the representative color of the cluster 700 is the ground color density. That is, the ground color detection (step 330) has a feature that the ground color of the paper can be detected without being influenced by the printed ground color.

  Returning to FIG. 3, after Yes in step 310 (after processing all the regions), the image correction unit 152 uses the ground color obtained in each partial region, and the density of the ground color of each pixel of the input image. Is created (step 340). Here, an example of the ground color density map creation process will be described with reference to FIGS. 10 and 11.

  The purpose of this process is to create the ground color density map (after interpolation) shown in FIG. 11 (hereinafter also simply referred to as “ground color density map”) from the ground color density map (before interpolation) shown in FIG. To do. That is, the following processing is performed on the ground color density map (before interpolation) in FIG. First, the brightness of the ground color obtained in each area is set as the ground color density of the center coordinates of each area. Next, the background color density of the pixel (for example, α) existing inside the four center coordinates (for example, A, B, C, D) forming the smallest rectangle is changed to the background color density of the four center coordinates. It can be determined by linear interpolation based on it. By executing this process on the entire image, the ground color density map (after interpolation) shown in FIG. 11 can be created. The ground color density map (after interpolation) shown in FIG. 11 is an image in which the shadow at the upper center is dark and the overall shade is smooth, that is, an image in which an actual shadow is reproduced with high accuracy.

  As another method, an approximate curved surface of the ground color density of the entire image can be obtained from the ground color density of the center coordinates of the region by polynomial approximation. Furthermore, as another method, instead of using the brightness of the ground color obtained in each area as the center coordinates of each area, linear interpolation is performed as the ground color density at the coordinates of the pixel closest to that color in the area. Alternatively, a ground color density map can be created by obtaining an approximate curved surface.

Returning to FIG. 3, after step 340, the image correction unit 152 determines whether or not pixel value correction is necessary based on the ground color density map (step 360), pixel value correction (step 370), and correction value set ( The process of (recording) (step 380) is performed for all the pixels (“No in step 350 → step 360 → (step 370 →) step 380”).
In the determination of necessity of pixel value correction (step 360), for example, if the luminance value of the ground color density map is equal to or smaller than a predetermined reference value, it is determined that pixel value correction is necessary (Yes).

  In the pixel value correction (step 370), correction is performed to remove the shadow from each pixel value in the image, assuming that the darker the background color, the darker the shadow. Specifically, the corrected RGB value is obtained from the luminance value of the ground color density map and the RGB value of the input image. As an example of the correction method, linear contrast correction will be described.

The linear contrast correction is a process for enlarging the difference between the maximum value and the minimum value of the brightness (luminance value) of the pixels in the processing area in the image. For example, if the maximum input value is Xmax and the corrected maximum value is 255, the input of 0 to Xmax is expanded to 0 to 255. As an example of linear contrast correction, there is a method of correcting each value of RGB of a pixel using the following equation (2). In Equation (2), X is the RGB value of the pixel of the input image, Xc is the RGB value of the pixel of the corrected image, Ys is the luminance value (density) of the ground color of the pixel, and α is the correction parameter. Expression (2) corrects the RGB value of the pixel of the input image according to the value of α · Ys. Further, min () means that the smallest one of a plurality of values in () is selected. That is, the maximum correction value is 255.
Xc = min (255 × X / (α · Ys), 255) (2)

The correction parameter α is a parameter for absorbing ground color variations, and takes a value of 0 or more and 1 or less. As the value of α is smaller, the corrected image becomes brighter. The reason for using α is to reduce the influence of variations in ground color in the cluster. Since the ground color in the cluster varies, if the representative color of the cluster is used as it is for correction, even if it is determined to be in the cluster, the shadow remains light in pixels where the ground color density is lower than the representative color. There is. When α is set to a value less than 1, the pixel value can be corrected based on a value lower than the representative color, so that even a ground color that is uneven can be corrected without leaving a shadow.
Note that the contrast correction may be performed by non-linear conversion according to the luminance value or ground color density of the pixel of the input image.

  Finally, in the pixel value set (step 380), the pixel value corrected in step 370 is stored as the pixel value of the corrected image. The pixel that is not corrected stores the pixel value of the input image.

  As described above, according to the image correction apparatus 100 of the first embodiment, dynamic shadows in a color image obtained by imaging a subject can be appropriately removed mainly by color clustering and a ground color density map. it can. As a result, it is possible to generate an image having no luminance unevenness without calibration each time the illumination condition changes, and it is possible to improve usability.

(Second Embodiment)
In the second embodiment, another implementation of the ground color detection (step 330) in FIG. 3 in the first embodiment will be described. In the ground color detection process of the second embodiment, first, color clustering is performed on partial areas, ground colors are detected, color differences or luminance differences between adjacent partial areas are evaluated, and the focused partial areas When the color difference or the luminance difference between two or more adjacent partial areas is larger than a predetermined value, the detected ground color is discarded, and a value obtained by interpolation processing from the adjacent partial areas is calculated. The ground color of the partial area is used.

  In the method of the second embodiment, as in the partial area 470 of FIG. 4, the ground color (white) area of the paper is not included in the partial area, and only the ground color of the ground color area 430 whose ground color is green is used. The purpose is to solve the problem that excessive correction is applied. The reason why excessive correction is applied is that the brightness value of the printed ground color area is detected as the ground color and is lower than the actual brightness of the ground color.

  As another implementation form of the ground color detection (step 330) in FIG. 3 in the first embodiment, the brightness of the representative color of the cluster with the highest frequency is used as the ground color density from the color clustering result in the partial region. Is possible.

(Third embodiment)
In the third embodiment, another implementation of the pixel value correction (step 370) in FIG. 3 in the first embodiment will be described. In the third embodiment, the character color is detected from the color clustering result, and the pixel value correction for the character color is different from the other pixel value correction. Specifically, the pixel in the cluster determined to be the character color is processed so that the correction amount is made smaller than other pixel value correction amounts or not corrected.

  The purpose of this processing is to prevent the luminance value of the character pixels from being too high after correction. This is because if the luminance value of the character pixel is excessively high, the character is faded by the binarization process, which adversely affects the character recognition process. As an example of the determination of the character color cluster, it can be performed based on whether the saturation (S), which is an index of color vividness, and the brightness (V), which is an index of brightness, satisfy a criterion. This is because the color of characters is generally black or a color close to black and has low saturation and lightness.

(Fourth embodiment)
In the fourth embodiment, another implementation of the ground color detection (step 330) of FIG. 3 in the first embodiment will be described.
In the fourth embodiment, first, the background color is detected after color clustering in the partial region. When it is determined that the luminance value of the detected ground color is equal to the ground color of the paper, the ground color B0 of the paper and the representative color Bi of the cluster whose pixel frequency (number) is equal to or greater than a predetermined reference. All (i = 1 to n) are stored. The reason why the representative color of the cluster whose pixel frequency (number) is less than the reference is not stored is to remove noise. The color Bi stored here can be assumed to be a color obtained by capturing the color printed or described on the form in a situation where there is no shadow. Specifically, for example, B0 is white with no shadow, B1 is black, B2 is red, and B3 is green.

  Next, the ground color Bk is detected in another partial area. If the brightness of Bk is low, there is a shadow in this area, and it can be determined that Bk is not the paper ground color, Bk is compared with B0 to Bn, and the color closest to Bk is obtained. For this comparison, hue (H), brightness (V) or the like, which is an index of the type of color, can be used. It is considered that the brightness of Bk is lower than the actual color due to the shadow. Therefore, as an example of a determination criterion, when the brightness of Bk is lower than the brightness of Bi and the hue difference between Bk and Bi is within the reference, it is determined that Bk is a color when the color of Bi is darkened by a shadow. . In addition to brightness (V), brightness (Y), HSL color space brightness (L), or the like may be used as a brightness criterion.

  Specifically, for example, when Bk is actually red and shaded, it can be determined that Bk is red because the hue difference from B2 is within the reference, and the brightness or luminance of Bk and B2 is determined. By comparing, the degree of shadow is known.

In such a case, the ground color density Ys used in Expression (2) does not use the ground color density Yk of the ground color Bk of the partial area as it is, but uses the brightness value Yi of Bi and the brightness value Y0 of B0. Then, it can be obtained by the following equation (3).
Ys = (Y0 / Yi) Yk (3)

  Expression (3) indicates that the luminance of the ground color of the paper is obtained from the luminance of the ground color of the detected colored region. Thereby, even when the ground color detected from the partial area is not the ground color of the paper, the luminance can be corrected appropriately.

(Fifth embodiment)
In the fifth embodiment, an example of an image processing apparatus capable of performing not only image correction but also character recognition will be described. FIG. 8 is an example of a configuration diagram illustrating an image processing apparatus 800 having a function of the image correction apparatus 100 according to any one of the first to fourth embodiments.

  As shown in FIG. 8, the image processing apparatus 800 has the same configuration as that of the image correction apparatus 100 shown in FIG. 1 except that a character recognition unit 811 is added to the control unit 810, and thus redundant description is omitted as appropriate. To do. The character recognition unit 811 performs character recognition processing from an image generated by image correction.

  For example, the character recognition unit 811 recognizes “1”, “2”, “3”, “4”, and “5” as characters (numbers) from the image including the ground color region 420 of FIG. Among the main processes of the character recognition unit 811, the process of detecting the position of the character can be cut out from the image using the form definition information in which the position of the character on the form is specified in advance. It is also possible to detect characters by detecting a frame in the image. The process for recognizing individual characters can use an existing method of pattern recognition. Note that image binarization processing may be executed as character recognition preprocessing.

(Sixth embodiment)
In the sixth embodiment, a branch office window terminal device installed at a bank counter of a financial institution using the image correction apparatus 100 according to any one of the first to fourth embodiments will be described. As shown in FIG. 9, the sales office window terminal device 900 includes an operation unit 110 (similar to FIG. 1), a display unit 120 (similar to FIG. 1), and an imaging unit 130 such as a stand-type scanner (similar to FIG. 1). ), A cash deposit / withdrawal unit 910 that handles cash for exchange with the user, a passbook printer unit 920 that prints transaction contents on a passbook, a printer unit 930 that prints on receipts, a terminal storage unit 940, and the like And a terminal control unit 950 for controlling. The terminal control unit 950 corresponds to the control unit 150 in FIG. 1, and the terminal storage unit 940 corresponds to the storage unit 140 in FIG.

  Hereinafter, a business flow using the sales office window terminal device 900 will be described. The sales office window terminal device 900 is a display unit 120, in which an operator who receives a transaction form for each transaction item from a user inputs a screen number for calling a transaction data input screen corresponding to the form from the operation unit 110. The target transaction is executed by inputting the entry data on the form such as the transaction amount into each input field of the transaction screen displayed on the terminal screen.

  Further, the terminal control unit 950 performs a correction for removing a shadow on an image of a specific part of the transaction form collected by the stand type scanner that is the imaging unit 130, and an OCR (Optical Character Recognition) function based on the corrected image. The description is read out, and a transaction confirmation screen is displayed on the display unit 120. In this case, the transaction details confirmation screen displays the contents of each data item read from the form by OCR, for example, the store number, account number, deposit amount, etc. Are transmitted to the management apparatus as recording data.

  As described above, according to the sales office window terminal device 900, even when a dynamic shadow due to illumination variation or the like occurs on the image collected by the imaging unit 130, the shadow can be corrected to be removed, and the OCR can be corrected. The processing accuracy can be improved.

DESCRIPTION OF SYMBOLS 100 Image correction apparatus 110 Operation part 120 Display part 130 Image pick-up part 140 Storage part 150 Control part 151 Image pick-up process part 152 Image correction part 200 Stand type scanner 210 Camera head 220 Column 230 Manuscript stand 240 Manuscript 800 Image processing apparatus 810 Control part 811 Character Recognition unit 900 Sales office window terminal device 910 Cash deposit / withdrawal unit 920 Passbook printer unit 930 Printer unit 940 Terminal storage unit 950 Terminal control unit

Claims (7)

An image correction apparatus that performs correction to remove a shadow from a color image obtained by imaging a subject with a non-contact scanner,
A storage unit for storing predetermined determination criteria for detecting the background color of the subject from the color image and a plurality of representative colors;
After dividing the color image into a plurality of partial areas, for each of the partial areas,
Perform color clustering, which is a process of clustering each pixel in the color space,
Determined representative color of each cluster using the results of the color clustering by the predetermined criteria to detect any of the determined representative color as a background color,
Based on the ground color of each of the detected partial areas , create a ground color density map showing the ground color density of the entire image area corresponding to the subject,
Using the background color density map, the shadow is darker as the background color is darker, and correction is performed to remove the shadow from each pixel value in the color image,
An image correction unit that stores the corrected pixel value in the storage unit;
An image correction apparatus comprising: Wherein the predetermined criterion, of the representative color of each cluster obtained for the partial region image according to claim 1, characterized in that it is to the highest color intensity and background color of the partial region Correction device. The image correction unit
For each pre-Symbol subregion, when detecting the background color from the above color clustering,
Evaluate the color difference or brightness difference between the ground colors of the adjacent partial areas,
If the color difference or the luminance difference between two or more adjacent partial areas is larger than a predetermined value for the focused partial area, the detected ground color is discarded and interpolation processing is performed from the adjacent partial areas. The image correction apparatus according to claim 1, wherein the value obtained in step 1 is used as a ground color of the focused partial area. The image correction unit
When performing correction to remove a shadow from each pixel value in the color image,
Detect text color from the result of the color clustering using the saturation or brightness index,
The pixel value correction of the character color is performed smaller than the pixel value correction of a pixel that has not been determined to be a character color pixel , or is not performed at all. The image correction apparatus described in 1. The image correction unit
When performing correction to remove a shadow from each pixel value in the color image,
A plurality of representative colors in the cluster detected in the partial area determined to have no shadow from the luminance value of the representative color are stored in the storage unit,
The ground color detected in the partial area determined to have a shadow from the luminance value of the representative color is compared with a plurality of representative colors of the stored cluster, and the detected ground color is determined from an index of hue or brightness. Select the representative color closest to the color,
Based on the ground color of the partial area determined to have no shadow, the selected color, and the ground color of the detected partial area, shadows are detected from the pixel values in the detected partial area. The image correction apparatus according to claim 1, wherein correction for removing the image is performed. An image correction apparatus that performs correction to remove a shadow from a color image obtained by imaging a subject with a non-contact scanner,
A storage unit for storing predetermined determination criteria for detecting the background color of the subject from the color image and a plurality of representative colors;
After dividing the color image into a plurality of partial areas, for each of the partial areas,
Perform color clustering, which is a process of clustering each pixel in the color space,
Determined representative color of each cluster using the results of the color clustering by the predetermined criteria to detect any of the determined representative color as a background color,
Based on the ground color of each of the detected partial areas , create a ground color density map showing the ground color density of the entire image area corresponding to the subject,
Using the background color density map, the shadow is darker as the background color is darker, and correction is performed to remove the shadow from each pixel value in the color image,
An image correction unit that stores the corrected pixel value in the storage unit;
A character recognition unit that performs character recognition based on pixel values stored in the storage unit;
An image correction apparatus comprising: An image correction method by an image correction apparatus that performs correction to remove a shadow from a color image obtained by imaging a subject with a non-contact scanner,
The image correction apparatus includes the storage unit that stores the color image and a predetermined criterion for detecting the ground color of the subject from a plurality of representative colors, and an image correction unit.
The image correction unit
After dividing the color image into a plurality of partial areas, for each of the partial areas,
Perform color clustering, which is a process of clustering each pixel in the color space,
Determined representative color of each cluster using the results of the color clustering by the predetermined criteria to detect any of the determined representative color as a background color,
Based on the ground color of each of the detected partial areas , create a ground color density map showing the ground color density of the entire image area corresponding to the subject,
Using the background color density map, the shadow is darker as the background color is darker, and correction is performed to remove the shadow from each pixel value in the color image,
An image correction method, wherein the corrected pixel value is stored in the storage unit.