JP2002183661A - Character recognizing device and image processor and image recognizing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processor capable of highly accurately recognizing a character. SOLUTION: The image processor is provided with a color separating means 2 for color separation of a part corresponding to a seal color and a color part of a writing-in character to input image data, a determining means for determining cancellation to a character writing part of image data of the color part of a write-in character separated by this color separating means, and a means 7 for outputting the existence of the seal color to the part by taking out a position part of the image data of the seal color, corresponding to a position of a cancellation part, when the image data is determined as the cancellation part by this determining means.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus applied to a character recognition device used for recognizing characters printed on a slip or the like.

[0002]

2. Description of the Related Art Conventionally, in the case of recognizing characters printed on a tabular form in which the arrangement of characters is separated by ruled lines and blank spaces, such as a slip, when inputting an image,
The colors constituting the form part of the form called drop-out colors are dropped by an optical filter process, the form part is deleted from the original image to be recognized, and image input is performed to perform character recognition.

[0003] For this reason, the prior art cannot handle a form having a color different from the drop-out color, and of course, cannot handle a form having a background color other than white. Further, it was not possible to handle multi-colored prints or printouts in which the written characters were drop-out colors, and of course, it was not possible to read those in which the colors of the written characters were not specified.

To solve this problem, Japanese Patent Application Laid-Open No. 61-15285 has proposed a character recognition device which can handle a document printed in a color other than a drop-out color.
Since the publication does not include a color separation technique, the separation of a character frame and a character is performed not by a color separation technique but by only a technique of separating a character frame and a character. Therefore, there is a limit in recognition accuracy.

On the other hand, JP-A-62-5485, JP-A-62-154181, JP-A-2-67689, JP-A-2-135584, JP-A-2-12
JP-A-3488, JP-A-3-14077, JP-A-3-223987, JP-A-4-39789, JP-A-4-160486, JP-A-4-354
Japanese Patent Application Publication No. 083 discloses an apparatus for performing color separation of an input image by a method set in advance and performing character recognition using the result.

[0006]

However, according to these character recognition devices, if the restrictions on the form color and the entry color, the printing method of the form form and the character portion, the character entry method, and the like are relaxed, for example, In some cases, characters cannot be recognized accurately, for example, when characters are randomly entered in a column, and it is not possible to cope with a case where color misregistration occurs.

In these conventional character recognition devices, in an operation mode in which drop-out color cannot be used, an image is read as color data, and some color separation processing is performed. Recognition is only proposed, and no solution is actually presented for various problems that must be considered when configuring a character recognition system.

For this reason, the prior art alone cannot be said to be effective as a character recognition means when an actual form reading operation mode as described above is actually implemented. As an example, for a form in which a part printed in red characters in a red frame and a part written in black characters in a red frame coexist, it is a problem how to read concretely, Conventionally, these problems cannot be solved.

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and has as its object to provide an image processing apparatus capable of realizing highly accurate character recognition.

[0010]

SUMMARY OF THE INVENTION According to the present invention, provisional color separation is performed on input image data, and geometric information of pixels constituting each provisionally color-separated portion, The final color separation is determined by using the density information of the part and the background together.

According to the present invention, the format information is prepared in advance with information on the definition of a partial area and color separation performed for each area, and color separation in a predetermined area is performed based on the information. ing.

According to the present invention, a specific color is separated from input image data of a form, format information corresponding to the separated specific color is selected, and recognition processing for the selected format information is executed. I have to.

According to the present invention, a plurality of representative colors extracted from image data are run-length-represented, a pointer table including pointers pointing to the run-length-represented data is generated, and a pointer of the pointer table is generated. Allows a specific data portion to be referenced.

Further, according to the present invention, each processing unit is prepared corresponding to color information, shading information, and binary information, which are color representation formats of an image, and a processing unit corresponding to the color representation format of an input image is prepared. I use it.

Further, the present invention has at least one of a color image buffer in which an input image is stored as it is or color-separated, and a gray image buffer which is obtained from the color image or stores a directly input gray image, And a binary image buffer for storing a binary image generated from a gray image or a color image. In the processing for the binary image, when rechecking a specific portion of the input image, the color image buffer And a corresponding portion of at least one of the specific portions of the gray image buffer.

According to the present invention, the input image data is color-separated,
The part corresponding to the color of the seal and the color part of the entered characters are divided, and a cancellation determination is performed on the image data of the entry color. If it is determined that the part is the canceled part, the seal corresponding to the position of the canceled part It has a double line cancellation processing function for extracting the position portion of the color image data, checking whether or not a seal color exists in that portion, and outputting the presence or absence of the presence.

Further, according to the present invention, a form image of a form is extracted from input image data, characters on the form are recognized, at least the position of the character is stored, and the color of the form image of the form is stored. When a form image is displayed as an image in a color similar to the above, the recognized character is displayed so as to overlap the position where the character existed.

Further, according to the present invention, a correction parameter necessary for correction is added to input image data as attribute data, and thereafter, the correction parameter is reproduced based on the attribute data, and the image data is corrected by the correction parameter. Image processing is performed on the corrected image data.

[Operation] As a result, according to the present invention, a form color and an entry color, a form form part and a character part printing method,
Even if the restrictions on the character entry method are loosely used, or if the entry characters are written in the entry column messy,
High-precision character recognition can be realized.

[0020]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 shows a schematic configuration when the embodiment is applied to a character recognition device. In the figure, reference numeral 1 denotes an image input unit to which image data is input. In this case, as the input image data, for example, R
Each of the three colors GB (red, green, and blue), each of which is represented by a multi-value, is used. Of course, other expressions may be used.

The image input unit 1 performs, for example, color shift correction processing, color bias correction processing, and color conversion.

The output from the image input unit 1 is sent to the color separation unit 2. The color separation unit 2 separates and extracts a color image of a specific color portion from the input image data, and separates the separated color portions into respective color image buffers 31, 32,.
n.

Each color image buffer 31, 3 of the buffer 3
The outputs from 2, 3n are sent to the image processing unit 4. The image processing unit 4 performs a run-length process described later.

The output from the image processing unit 4 is
It is sent to the character cutout unit 5. The character cutout unit 5 performs character cutout processing from the image data read from the buffer 3.

The result of the cut-out processing in the character cut-out unit 5 is sent to the character recognition unit 6. Character recognition unit 6
Is for performing recognition processing on characters cut out by the character cutout unit 5. The result recognized by the character recognition unit 6 is sent to the read result correction unit 7, corrected by the operator, and output as the final recognition result.

In such a configuration, image data is input to the image input unit 1. In this case, the input image data is, for example, one of three colors of RGB (red, green, and blue) and is represented by multi-values.

Then, the image input unit 1 performs a process of correcting a shift with respect to the input image, a process of correcting a color bias, a color conversion, and the like.

Here, the color misregistration correction may be, for example, a parameter describing a shift amount for each color of RGB as a color misregistration correction parameter. Is achieved.

The color bias correction process is performed, for example, by using RG
This is performed by converting the color density for each color B. Here, as the method of converting the color density, a method of converting the density of a single color from the density of one color called gamma conversion may be employed to implement the color bias correction processing,
A function that determines the density of each color from the density of all RGB colors may be prepared. The color bias correction parameter here may be numerical data for determining those functions, or may be the conversion function itself.

The present apparatus holds these color shift correction parameters and color shift correction parameters, and performs color shift correction and color shift correction processing based on these parameters. That is, during actual character recognition processing and image processing work such as character recognition, feature extraction, line segment extraction, and binarization processing, unrecognition and extraction failure increased, and some abnormalities and contradictions occurred. Sometimes color shift correction not depending on correction parameters,
A new correction parameter is obtained, and the parameter is updated by newly rewriting the parameter when the result is different from the correction parameter.

This updating may be performed periodically, or may be performed according to an operator's instruction. Further, a plurality of color correction (correction) parameters may be prepared so as to correspond to a plurality of image input units 1. In this case, the color correction using the respective color correction (correction) parameters and the update of the parameters are performed for each image input unit 1.

The color misregistration correction parameter is obtained by color analysis based on a black portion in a form or the like. The black portion may be automatically extracted from the input image, or its location and size may be determined in accordance with preset format information.

When a pattern having a predetermined shape is set on a form, a pattern or a mark for correcting color misregistration, for example, a cross, is set on a specific portion of the form, and based on the set pattern or mark. Processing will be performed. Information such as the position and shape of these patterns and marks may be registered in the format information. Also, this pattern portion can be shared with other uses as a character frame for describing a form number.

On the other hand, the color misregistration analysis for correcting color misregistration is performed by measuring how many complementary colors exist in the upper, lower, left and right sides of the black portion.
Such measurement may be performed at a plurality of locations, and an average of the results may be taken as a color shift amount common to all screens, or may be regarded as a color shift amount at each location. The latter case can cope with the case where the color shift is not uniform over the entire screen.

FIG. 2 illustrates a method of measuring color misregistration. In this case, as shown in FIG.
If there is a double bleed, that is, a double bleed, the breaks of each color are represented by A1 to A6. In this case, A1
The distance from A1 to A2, provided that the area of A2 and the area of A4 to A5 are in complementary colors, and the area of A2 to A3 and the area of A5 to A6 are in complementary colors.
The distance from A1 to A3 is determined as the amount of color shift between green and blue with respect to the red signal.

Here, the relationship between the complementary colors means that the sum of the two is outside the background colors A1 to A6, in this example, white or close to it. This concept can be similarly applied even when the central areas A3 to A4 have disappeared.

In this case, a single overhang can be applied instead of a double overhang. In the case of a single color, the shift amounts of the two colors may be the same. This example shows an example in which the colors are shifted in the order of RGB. However, the order in which the colors are shifted can be determined from the color of each area.

The above concept can be applied to a non-black portion of a form. In this case, a color misregistration correction parameter is determined based on a change in color between adjacent portions. That is, the RGB density values of the background color are R0, G
Assuming that the color of the target portion is R1, G1, and B1, the color change at the place where the color shift occurs can be represented as shown in FIG. And the amount of color misregistration can be determined.

In this case, the areas A1 to A2 and A4 to A4
If the sum of the area of 5 is taken, it becomes the sum of the color of the area of A3 to A4 and the color of the background.
If the same can be said for the areas 2 to A3 and the areas A5 to A6, the distance between A1 and A2, A
The distance between A1 and A3 is determined as the amount of color shift between green and blue with respect to the red signal. Further, even if the areas A3 to A4 have disappeared or if the area does not protrude double but protrude single, R0, G0, B0,
Since R1, G1, and B1 can be determined, the amount of color misregistration can also be determined using the above-described concept.

On the other hand, a color printed mark as shown in FIG. 4 is detected, and the amount of color misregistration can be detected by examining the misalignment of each of the red, green, and blue colors therein.

Further, it is also possible to execute the color bias correction processing from each color of such a mark. That is, it is possible to determine a parameter or a function for correcting color bias based on the difference between the color information of the mark obtained from the image input unit 1 and the color information that should be expected.

Further, a pattern for correcting color misregistration is set at a specific portion on a scanned surface of a scanner for generating image data to be input to the image input section 1, and based on a pattern in an input image to be created from this. Color shift correction or color bias correction may be performed.

FIGS. 5A and 5B show a method for obtaining further different color shift amounts. In this method, matching is performed while shifting the waveform of each color in a portion where the density value of each color of red, green, and blue is changing, that is, overlapping is performed,
The shift amount of each color is obtained at the position where they overlap most. The matching method here can use a matching technique often used in the field of pattern recognition.

The operation of correcting the color misregistration is performed only on the changed portion of each color as shown in FIG. 5, and the portions which are not affected by the color misregistration, such as the background portion, the printed portion and the described portion, are subjected to the color misregistration processing. May not be applied. This is because in these portions, the result does not change even if each color is shifted. This speeds up the correction of color misregistration. The function of enlarging and displaying a portion where a color shift of an input image can be clearly recognized or a specific position can make it easier for an operator to check the color shift state.

In the color misregistration correction, it is possible to shift the input image in units smaller than one pixel by using, for example, interpolation. In this case, the density value at the position to be moved is determined by interpolation calculation from the density values of the pixels near the position.

Note that the above description can be applied to a case where the color blur is different. FIG. 6 shows an example in which red appears at both ends of black. Such a case indicates that the focus of green and blue is out of focus with respect to red and blurred.

As described above, when the same color is present on both sides instead of the complementary color, it is possible to obtain a blur correction parameter, whereby the degree of blur of each color can be made uniform.

The above-described various correction and correction operations are not performed in advance, but are performed in consideration of the correction and correction parameters in a character extraction process and a character recognition process described later. You may do.

The output from the image input unit 1 which has undergone the color shift correction and the color bias correction in this manner is sent to the color separation unit 2.

Next, a method of color separation in the color separation section 2 will be described. Here, the form color registration function will be described.
When you fill out a form (completed or unfilled) in R,
A color analysis is performed from the image to find out what color the form has, that is, what part has what color, and based on that, the color information is registered in the format information. The registration data is simply the number of color types and color data (RGB)
It is possible to provide a means for registering from the above to the actual form image.

At this time, of course, the operator can interactively modify the registered contents. For example, for a form image input as a test, color extraction (erasing) of various colors is performed by trial and error, and a color that allows a printed portion of the form to be extracted (erased) well is registered in the format information. Means.

Further, an artificial color data or an input image of a form is displayed on the screen, and by specifying a certain position on the screen, RGB, VSH and color names of the position are output.
It is also possible to make it easy for the operator to determine the registered color.

For such processing, it is convenient if the zoom function allows the operator to enlarge or reduce a part of the image.

A process for separating and extracting a specific color is executed based on the format information thus created.

Color information registered in the format information (including the background color and the entry color, each of which may have a plurality of possibilities)
, The part to be erased is removed, and the part to be recognized is extracted and recognized.

In this case, it is possible to configure such that the part to be erased is automatically determined only from the color information obtained from the input image without depending on the format information. For example, in a simple example, only a specific color can be erased, or the color of a form portion can be automatically determined automatically from the linearity of a form of a form and the color. The same applies to the part to be recognized.

The color separation can be performed by converting RGB into lightness, saturation, and hue (VSH) and based on the conversion. It is also effective to define a distance in an RGB or VSH space and separate colors based on the distance.

The method of defining color names and separating colors based on them is a powerful method in consideration of the man-machine interface. In this case, the definition in the format information is also performed by the color name.

In such a system, the name of each color, RG
Color expression data such as B and VSH are registered in association with each other. Therefore, the color name registered in the format information is R
When color expression data such as RGB or VSH is used by being converted to color expression data such as GB or VSH, and conversely, when color expression data such as RGB or VSH is presented to the operator, it is inversely converted into a color name closest to that and notified to the operator. You.

Next, a general method of color separation will be described. Here, an example of a gray image will be described for simplicity.
FIG. 7A shows an example in which the frequency distribution of density values of a gray image is obtained.
(B) and (c). In this case, 0 is black, 1 is white, and the middle part is dark gray to light gray. The distribution of A is obtained by taking the frequency distribution of the image shown in FIG. In this figure, the number "5" (A
This is an example in which a light gray straight line (part B) penetrates the center of the part (part B). These frequency distributions are shown in FIG.
A and B of FIG. At this time, a part b of the straight line part is dark, and a part a of the number “5” is bright corresponding to b of the frequency distribution, and corresponds to the frequency distribution a. Thus, the entire frequency distribution becomes C. When density separation corresponding to color separation is performed from this frequency distribution, the frequency is divided by the density value m and divided into X and Y areas.

However, when an image is separated by this separation method, the portions a and b remain as noise as shown in FIGS. 9A and 9B. Also, the straight line is divided by the number “5”.

The data is not only density information but also RGB
In the case of the color information of the three colors
What is a dimension can be extended to three dimensions. What is necessary is just to consider the color of the area | region in a three-dimensional space as the same color, and to carry out color separation.

However, if color separation is performed only by dividing the area in the color space having such a density, correct color separation cannot be performed as described above. In this case, unless the geometrical information of each pixel, that is, the phase information is used together with the density information, such as that the part A is a straight line and the part B is a numeral “5”, correct color separation is logically possible. What is impossible is clear from the above explanation.

This will be described in more detail.
It is assumed that the result shown in FIG. Now, for each block a, x, y, z, w, check that for each set of them they are part of a straight line. As a result, for example, since the density value is dark in an area between x and y, there may be some black line between them, and x and y may be part of a straight line. There is a nature. The same is true for y and z, z and w, and the linearity is positive for x, y, z and w as a whole. These are connected by extending the outline of each block to obtain one straight line as shown in FIG.

Also, for example, a and y cannot be connected because the extension of the outline does not match or a background color exists between them. Therefore, a is an isolated block and is removed as noise. Conversely, the dent of b is found by tracing the contour, but since the density of the dent is dark, but the contour coincides with the contour when the entire xyz is regarded as a straight line, b is determined to be a part of a straight line, and finally a straight line as shown in FIG. 11 can be obtained.

The same applies to the numeral “5” shown in FIG. 9B, and the numeral “5” can be correctly extracted from the information on the likelihood of a curve and the shape of the character.

That is, in the present invention, after provisional color separation, provisional character cutout processing, character recognition processing, line segment extraction processing, form structure understanding processing, etc. are performed, and the final result is obtained by using those results. It is to determine a color separation result. In other words, if there is a place where it is judged that color separation is suspicious during the processing of recognition, extraction, understanding, etc. or as a result, feedback is performed, and the separation processing is redone while using the result of color separation of the relevant part. It is configured to execute.

As another example, for example, if a lot of straight lines are included, and if necessary, their arrangement is checked to find out the form part of the form (unfilled form screen).
Can be automatically determined as the color of the form part of the form.

In the case where chunks about the size of a character are continuous, those colors can be determined as the writing color and the printing color of the character.

As described above, according to the present invention, even when there is no color information or format information itself registered in the format information, a plurality of representative colors constituting the form are automatically separated by analyzing the color of the input form. It is possible to Of course, when the color information and the format information are defined in the format information, it goes without saying that the above object can be achieved more easily. Next, FIG. 12 shows an example in which a noise d exists near a straight line composed of regions a, b, and c. And the colors of a and c are C1,
It is assumed that the colors of the regions b and d are C2, and C2 is an intermediate color between C1 and the background color.

In such an image, an arbitrary point is used as a region seed, and the region is gradually expanded starting from there. The area is enlarged by taking in pixels that can be regarded as the same color as the color of the area and connected to the area to the area. By repeating this process until it does not belong to any region, four regions a, b, c, and d shown in FIG.

Such a region extraction method is a region segmentation method known in the field of image processing, and is one of methods called a region growing method. In addition, various methods such as split merge and relaxation are possible.

Next, for each area portion, b is determined to be a part of a straight line, that is, b from the determination of linearity in the same manner as described above.
Is separated as C1, and the portion of d is separated as noise, that is, as the color of C2.

Generally speaking, according to the present invention, representative colors of an area are determined by performing area segmentation, and the overall colors are determined by comprehensively judging them. Further, color separation is performed from the information on the representative colors as needed.

The correct color separation according to the present invention can also be performed by introducing information on the proximity of colors to the area segmentation. For example, if the determination that the colors are the same when the growth method is performed is relaxed, it can be set so that a, b, and c are detected as one region.

That is, when it is determined that the colors of two areas having many contour portions in contact with each other in the course of growth are close to each other, specifically, it is determined that C1 and C2 are close to each other.
These are merged into one large area. In this way, there is no color close to C2 that greatly touches d, or it is a background color, so it is ignored anyway as noise, while region b is correctly extracted as part of a straight line.

In this method, since the phase information of the pixel, that is, the geometrical relationship of each pixel is used in addition to the color information, the representative color is determined based on only the conventional statistical information of the color. This makes it possible to determine a representative color and perform color separation with higher accuracy than the method of performing color separation.

In order to separate and extract a specific color, the color information of the image and the one represented in the image, that is, a figure,
It is effective to use both shape information such as straight lines, curves, and characters. FIGS. 13A and 13B show an example thereof. Numeral 131 indicates that the middle of the blue straight line is changed to blue-green. If the blue color is simply separated from this, the straight line will be blurred as shown by 132. Further, if the blue-green portion is also regarded as blue because of this, for example, 13
In the case where blue-green dust is attached to the blue straight line as shown in FIG. 3, the image with the dust attached as shown at 134 is color-separated.

Therefore, in the present invention, first, shape analysis is performed on the blue portion, and it is recognized that the portions P and R are straight lines. Then, the portion Q which is regarded as a blurred portion is examined, and the P, Q, R Provided that the colors are close, and that P, Q, and R have a shape that can be regarded as a straight line as a whole,
Color separation is performed by regarding all of Q and R as a blue area. In this way, even in the case shown by 133, the dust area A is not erroneously separated into blue.

FIG. 14 shows how the blue line and the red line overlap. A1, C and A2 are blue lines, and B1, C and B2 are red lines. C is purple in the overlap area.

In this case, C cannot be extracted by extracting red or blue by the usual method. Also, if purple is always extracted when red is extracted, it may be inconvenient that a part that is truly purple may not be extracted.

Therefore, for example, in the case of removing blue, first,
Observe the area in contact with blue, and when there is a color that may be a mixture of colors other than the blue,
Further, the vicinity of the mixed color area is examined, and if there is an area where the other color forming the mixed color exists, the mixed color area is also extracted as a blue area.

At this time, the shapes of the blue areas A1 and A2 represent straight lines, the areas B1 and B2, which are the other colors forming the mixed colors, also represent straight lines, and the area C also has those lines. By checking that there is no inconsistency with the straight line, more accurate color separation can be achieved.

In this case, the shape check may be performed not only by checking a straight line but also by checking a curve, a figure, a character shape or certainty of a character.

FIG. 15 is another example. If area C is reddish black, color separation can be performed in substantially the same manner as above, but if the mixed color of red and black is black, x and z, y and w
, The color separation of red becomes possible. Further, even when the area B2 does not exist, red color separation can be performed by estimating a curve connecting x and z from the shape of the area B1.

The method of performing color separation by using both line width information and color information is effective. In this case, if the specification relating to color separation is described in the format information, information such as a line width threshold and a line pitch in the case of hatching are set in addition to information relating to the color of the object to be color separated. .

FIG. 16 shows a circular area 162 set at a certain portion on a line 161 observed on a form.
For example, there is a known method in which the black is examined in the N direction from the position of the center of gravity to determine the existence of black, and the length of the black pixel in the direction in which the continuous length is minimized is regarded as the line width. Have been. As another means, a method of obtaining an estimated value of the line width from the run length of the vertical and horizontal black pixels can be considered. In any case, a method of obtaining partial line width information in this manner is known.

Thus, it is possible to use both the fact that the part has a specific color or is not a specific color, and the fact that the part is a part that is smaller than the line width determined. Thus, effective color separation can be performed.

With the above method, for example, a character portion 171 having the same color and a wide line width can be extracted from a hatched area as shown in FIG. 17 on a form. At this time, even if the line width is narrow, the portion having a different color is, for example, 17 lines.
The portion like 2 can be left separately even after the hatched portion is erased.

By the way, the following are the main methods of specifying color separation. When separating a specific color, when separating colors other than a specific color, when separating achromatic colors, when separating chromatic colors, when separating white, when separating black, or a combination thereof.

When a specific color is separated, the determination is made mainly based on the value of the hue (H). In the case of separation of achromatic color and chromatic color, the determination is made mainly based on the value of the saturation (S). When separating white and black, the determination is made based on the magnitude of the lightness (V) from among those having the small values of the saturation (S).

From the result after the color separation, the most likely background color in the form is
For example, you can choose the most dominant representative color. That is, the color having the largest total area is selected from the representative colors. It is useful to add this result to the recognition result and output it. That is, the automatically estimated background color is added to the recognition result and output.

It is also useful to output the color of the recognized form part or the color of a specific part on the form.

The character code obtained as a result of recognition is output after adding the color of the character on the form, the background color of the position where the character is described, the color of the character frame surrounding the character, and the like. It is also useful to do so.

Further, the color information of the form part can be used for form discrimination together with the ruled line information. On the basis of the extracted color information, an alarm can be issued to the operator or a message to that effect can be displayed. Further, it is also possible to change the display of the correction screen based on the color information.

It is very important to change the color separation means for each partial area on the form.

First, the format information previously describes the definition of a partial area and whether to perform normal binarization processing or color separation processing within the area. It is assumed that the form shown in FIG. 18 is input at the time when the actual processing is performed. In this form, the area to be recognized is 181
182 is an area where only the image is taken out and the image data is to be outputted as it is. It is assumed that the format information is described as such. In this case, in the area 181, it is necessary to recognize the character by removing the preprinted portion indicating the character frame. On the other hand, in the area 182, preprinted portions such as character frames and item names must not be erased, so that their colors are not removed. In this case, if they are uniformly color-separated, a problem occurs.

The specific processing in this case is performed as follows. For the area 181, an image is input in color data based on the format information. On the other hand, with respect to 182, an image is input in binary. Regarding 182, a monochrome image may be obtained by calculating an average value of RGB with respect to the input color data, and the monochrome image may be processed to obtain a binary image. Next, the area 181 is determined based on the format information.
Then, the color of the character frame is separated and deleted. Then, character recognition is performed on this. For 182, the image data is output as it is.

By doing so, it is possible to limit the time-consuming color separation work to only a part of the form. Since ordinary binary processing can be executed in a short time, the overall processing time can be improved by limiting the color separation work to a part.

In the above example, whether or not to perform color separation is changed for each partial area. However, a method in which the color separation means is selected from a plurality of methods is also possible.

For example, in the area of the seal portion of the area 182, a color separation process set to perform red or orange color separation is started, and in the area 181, a process of separating black is performed. By registering the format information so as to be activated, character recognition and stamp extraction can be performed at high speed. In this way, conventional methods, such as separating all representative colors for all areas,
For example, processing can be performed at a higher speed than in the method disclosed in Japanese Patent Application Laid-Open No. 3-14077.

In the present invention, attached to these designations, there are designations such as character cutout means, character recognition means, and character recognition post-processing means in the area which allow a specific one to be selected from a plurality of methods. You may be able to do it.

A mechanism for color-separating a specific color in a specific portion of a form, determining whether the color exists in that portion, recognizing an object of that color if it exists, and outputting a recognition result. It is convenient. When it is determined that there is no recognition target object of the color, it is output to that effect. At this time, the color of the object described in that portion may be output. Of course, at this time, it may be configured to output only these determination results and color information without recognizing the recognition target.

At this time, it is possible to determine whether or not a seal is present in the relevant part and to determine what kind of the seal is by incorporating the seal determining unit into the apparatus, and to output the result as a recognition result. Can be. For example, the system can be configured to automatically notify or warn the operator when there is no seal or when a predetermined seal is not pressed. That is, based on the extracted seal stamp information, an alarm can be issued to the operator or the effect can be displayed. Further, it is also possible to change the display of the correction screen based on the seal information.

The seal determining unit determines, for example, that a target portion of the color is present from an image separated by a specific color, for example, red, and that the contour shape outside the portion is, for example, a circle or an ellipse. And the fact that the internal shape is complicated to some extent.

The similarity of the shape of the seal can be determined by a pattern matching process often used in pattern recognition and the like.

A result obtained by performing color separation of a form and a result obtained by subjecting the form to some processing may be displayed on a screen, and the processing may be advanced to the next processing after an operator's confirmation. In this case, if the operator is dissatisfied with the processing result, a mechanism that can directly correct the processing result may be provided, or a configuration may be adopted in which any parameter is changed and the processing can be performed again. .

When a color that is difficult to determine is found at the time of color separation, if necessary, it is displayed together with its surrounding state to prompt the operator to input a correct answer.
Thereafter, the color separation processing can be performed based on the correct answer data.

First, a portion corresponding to the entry column includes a portion that is color-separated into a first specific color, for example, black. If these are recognized as a character string, they are output as a recognition result. Similarly, if necessary, the same processing may be repeated up to the N-th specific color. If these specific colors are not the same as the form color of the form, character frames and entry fields are assumed to be erased in the color-separated image data. "Recognition method" is applied. Conversely, when they are the same, the “character recognition method with a frame” is applied. If it is assumed that nothing is present anywhere in the entry field, output a blank.

In addition, the following can be performed separately. If a portion separated as a color that is not the same as the color of the form of the form exists in the entry column and these are recognized as a character string, they are output. In this case, it is assumed that the character frame and the entry field have been erased, and the “frameless character recognition method” is applied to character recognition. If the existing area of the same color as the form is small in the entry field, the characters in the entry field are assumed to be described in the same color as the form, and the "character recognition method with a frame" is applied for character recognition. Is done. If it is assumed that nothing is present anywhere in the entry field, output a blank. Either of these processes may be performed first.

The "frameless character recognition system" performs character extraction and character recognition on the assumption that there is no character frame, and various systems are known.

The "character recognition method with a frame" is to perform character cutout and character recognition on the assumption that a character frame exists. There are two types of "character recognition method with character frame with known shape", in which the position and shape of the character frame are given in advance from format information and the like, and "character recognition method with character frame with unknown shape", which is not so. Either of these is adopted depending on the contents of the format information.

In each method, there are a "character recognition method with ruled line contact" and a "character recognition method without ruled line contact", and either method is selected from the actual pattern state.

These are executed according to the following procedure. First, a character frame estimation process is performed to determine a character frame position or a candidate thereof. Next, the character portion is estimated, and the degree of contact between the character and the character frame is examined. If there is little or no contact here, "character recognition without ruled line contact" is selected. Of course, it is also possible to configure so that the method is not divided between contact and non-contact. Next, character extraction and character recognition are performed based on the obtained ruled line position information.
If there is a ruled line contact, an appropriate character and ruled line separation process is performed.

As for the ruled line separating technique, the character extracting processing and the character recognizing processing accompanied with the contact character separating processing, those which are known as the prior art can be used even if they are not.

With such a configuration, it is possible to configure a recognition system with much higher accuracy than character recognition by the same method at all times.

Note that color separation and color shift correction can be performed simultaneously. At the same time, color deviation correction and correction processing for adjusting the degree of color blur can be performed. For example, in the case of separating a black portion, if the color misregistration correction parameter is known in advance, it is only necessary to simultaneously perform the shift of each color of RGB and the color separation (operation of extracting a specific color). Also, even when the color misregistration correction parameters are not determined in advance, if there are complementary colors on both sides of the black portion, it is assumed that the color misregistration is estimated from them and the RGB colors are shifted based on the result. What is necessary is just to separate the black part after having done.
Similar processing can be performed for other colors.

Next, the expression of colors in the image processing section 4 will be described. In the image processing unit 4, a plurality of representative colors obtained by color separation, that is, colors selected ignoring subtle color differences, or input image data, , A pointer is created for this one-line representation, a table consisting of this pointer row is created, and the data consisting of this pointer row and the representative color run-length representation is stored as image data. Then, using such an image expression as input, image processing or character recognition is performed based on the color expression.

First, character recognition and image processing can be performed by converting a part of the color run-length representation image into the original bitmap data. Unless a run length expression for each color or a pointer for each color is used, some data cannot be decoded, and the entire image must be decoded. There is also known a method of performing character recognition on an image expression in a format in which a table represented by a pointer and a black-and-white run-length data expression are combined.
As it was, color images could not be handled (U
SP4426731). If you want to convert a specific color part of a certain part into bitmap data, refer to the run-length data expressed in color from the pointer of the line corresponding to that part and convert the bitmap only for the specific color. Or, by referring to only the pointer of the specific color of the part and obtaining the bitmap data from the run length data indicated by the pointer, it is possible to quickly extract only the specific color of the specific part, Character recognition and image processing can be performed based on this. This makes it possible to perform processing much faster than performing the above processing by decoding the entire screen. Further, if processing is directly performed on the run-length expression without decoding, higher-speed processing becomes possible.

The representative colors appearing on the form are described in the format information in advance, and are extracted based on the information or automatically extracted otherwise. Express like.

In this case, assuming that a code C representing a representative color and a continuous length of the color or L is a coded version of the color, a set CL thereof is obtained, and an image of one line is obtained by the continuation. To express. FIG. 19 shows an example of coding. This is a colorization of the run-length expression.

Further, a method of performing run-length encoding for each representative color can also be used. In this case, 20 in FIG.
There is an expression method in which run-length expressions of respective representative colors are put together for each line as shown by 0, and an expression method in which run-length expressions of the entire screen are put together as indicated by 201 in FIG. Also, the method of setting a pointer that points to the head of the run-length expression of each line is described in FIG.
There are a method of collecting pointers of each representative color for each line as shown in FIG. 02 and a method of collecting pointers for all screens of each representative color as shown at 203 in FIG.

Further, each color of RGB is quantized, and each RGB is
Similar run-length coding can be performed for each time. These codings are realized by considering slight color differences between adjacent pixels as noise and ignoring them, and approximating them with representative colors or quantized densities.

Examples of directly performing image processing on a run-length representation image are disclosed in, for example, Japanese Patent Application Laid-Open Nos. 62-246138 and 62-307615. These are all image processing methods for black-and-white run-length data, but can be applied to run-length representations of each color. It is also possible to execute character recognition based on these image processing techniques.

For example, when detecting a long straight line of a specific color in a specific area, referring to a pointer indicating the specific area, length data representing a specific color of a portion including the area is obtained. A straight line can be detected by obtaining a long one included in the partial area and integrating adjacent long runs.

As another example, an example in which image processing is performed directly on an image expressed in color run length will be described below. An example of the image processing in this case is a color shift correction processing.

By performing image processing, character extraction, and character recognition based on these coded information, the processing speed can be increased as compared to, for example, processing on RGB images, and data representing images can be obtained. The amount can be compressed.

It is also possible to correct the color misregistration of the data represented by the color run length. For example, RG
When a run expression is created separately for B (203 in FIG. 20 (d)), shifting by one pixel can be easily realized only by changing the length of the first run of each row.

In the case of a shift of one pixel unit or less, the shift is realized by the following processing. First, the value of the pixel at the boundary of each run is checked, and the length of each run is converted from an integer to a real number based on the density value. The length of each run is changed so that the longer the density of the color at the boundary pixel is, the longer the run is, and the lower the density is, the shorter the run is. At this time, do not change the grand total of each run.

Next, the length of the first run is changed according to the shift amount. If the run expression of one line is performed from left to right, the first run is shortened by that amount when shifting left. When shifting to the right, make it longer. Then, the length of each run is converted to an integer so that the total sum of each run is not changed. with this,
A shift of the image on the run representation is realized.

For example, when an image is represented by a color code + run-length expression (201 in FIG. 20B), a continuation of three or four runs in each one-line run expression is examined.
The amount of shift and the amount of the run are judged based on the color of the run and whether or not the color and length of the run sandwiched between the runs of the continuous run may be caused by color misregistration. In the case of, the run is extinguished, and the length of both runs to be changed as a shift of each color is changed. Alternatively, each run is color-separated into, for example, RGB, and a color shift shift is performed instead of the expression 203 substantially in FIG. 20D, and the original expression (201 in FIG. 20B) is returned. Is also possible.

In any case, it is possible to shift each color of the color run expression by such means without directly returning to the original image.

By performing the above-described color separation in two stages, the processing can be speeded up. In the first stage, color separation is performed so as not to disregard a small color difference, a run-length expression of the first color is created based on the result, and then a second-length expression is generated based on the data of the run-length expression. Color separation. If necessary, a second color run-length data expression is created based on the result of the second color separation, and this is used as input data for image processing and character recognition processing.

Generally, color separation takes a long processing time.
If a simple and high-speed method is used for the first color separation to roughly perform color separation, and a second color separation requiring a long processing time is performed, a precise representative color can be determined. Thus, the separation process can be performed at a higher speed than when the second color separation process is directly performed on the original image.

As the second color separation means, each of the above-mentioned methods can be considered, and these are color separation means requiring processing time. On the other hand, Japanese Patent Application Laid-Open No. 4-160486 is suitable for the first color separation because it is simple and fast.

The RGB values are: lightness (V), saturation (C),
It is better to convert to hue (H) and perform color processing in these spaces, and it is more convenient to specify colors by these values. These conversions are realized, for example, as follows.

First, the lightness (V) can be defined as an average of RGB density values. Next, the saturation (C) can be defined based on the minimum density value. That is, the saturation takes a higher value as the minimum density is closer to zero. This is because, in this case, since the difference between the density values of the color having a high density value and the color having a low density value is large, a vivid color can be obtained. In order to define the hue (H), first, a numerical value corresponding to each color of RGB is determined.
For example, assume that R is 0, G is 128, B is 192, and B is R
R represents 256 when representing a color change leading to.
Next, a numerical value between the numerical value indicating the color of the maximum density value and the numerical value indicating the color of the next density value is determined by the ratio of those density values, and is set as the numerical value of the input color, that is, the hue.

Specifically, it can be obtained by the following procedure. r, g, and b are RGB concentration values, h, s, and v, respectively.
Is described as the HSV value, and the conversion procedure is described in C language as follows.

V = r + g + b; if (v = 0) ｛c = 0; h = 0;｝ else ｛r = (r * 255) / v; g = (g * 255) / v; b = (b * 255) / v; if (r> g) ｛if (g> b) ｛/ * r>g> b * / s = 255−b * 3; h = ((g−b) * 129) / (r + g −2 * b);｝ else if (b> r) ｛/ * b>r> g * / s = 255−b * 3; h = ((r−g) * 65) / (b + r−2 * b ) +192;｝ else ｛/ * r> = b> g * / s = 255−g * 3; h = ((r−g) * 65) / (r + b−2 * b) +192;｝｝ else ｛if (R> b) ｛/ * g> = r> b * / s = 255−b * 3; h = ((g−b) * 129) / (g + r−2 * b);｝ else if (b> g) ｛/ * b>g> = r * / s = 255−r * 3; h = ((b−r) * 65) / (b + g−2 * r) +128;｝ else if ((g> b) ‖ (b> r)) ｛/ * g> = b> = r * / s = 255−r * 3; h = ((b−r) * 65) / (g + b−2 * b) +128; ｝ Else ｛/ * g = b = r * / s = 0; h = 0;｝｝｝ if (h> 255) h = 0; if (s> = 128) s = ( s−128) / 4 + 224; else s = (s * 224) / 128; v = v / 3;) Such conversion is not limited to the above, but it is possible to obtain these values by table lookup. You may do it. That is, the RGB values are quantized, and each color is set to N
Expressed in R, NG and NB bits, the address space is NR +
An NG + NB bit table may be prepared, and the HSV value corresponding to the color represented by RGB may be stored in the address corresponding to the RGB value in this table.
The color representation of the image input from the image input unit 1 may be HSV.

In addition, these conversions may be performed simultaneously with the correction of the color deviation, or these conversions may be performed simultaneously with the correction of the color shift.

By the way, there are three types of image representations. Each pixel holds color information (C), and each pixel holds density information (G). (B).

The former is suitable for more advanced recognition processing, while the latter is suitable for high-speed processing. Assigning color data to portions that do not require advanced recognition processing or assigning binary data to portions that require advanced recognition processing is not desirable from the viewpoint of improving system performance. Therefore, it is important to introduce a mechanism for designating a data expression format for each part on a form. Specifically, an area designation and an image expression format in the area can be described in format information or the like.

The image input unit 1 to output any of the above CGB images is not limited to one type. Therefore, in the device of the present invention, a mechanism for character recognition corresponding to (C), a mechanism for character recognition corresponding to (G), and (B)
, At least a plurality of mechanisms for character recognition are provided, and processing is performed according to the type of input image.

FIG. 21 shows a configuration for embodying the above-described means. Outputs from the image input unit 211 are respectively captured into a color image buffer 212, a gray image buffer 213, and a binary image buffer 214, and Character recognition is performed by the recognition units 215 to 217, and these recognition results are output via the read result correction unit 218.

Each of the forms to be read has a CG
Since there is a choice as to which of B the image should be input, it is convenient if the format information or the like can specify the image expression of the image to be input. The image input device is requested to output the specified image expression according to the specification regarding the image expression of the format information. Since the image input unit 1 sends an image corresponding to the image, the present apparatus performs a process corresponding to the image.

However, if the image input unit 1 cannot respond to the request or does not make a request from the present apparatus, data of an already known image expression is not sent. Recognition processing or the like is performed according to the type of image expression data attached to the data or set in advance.

When the object of image input is an image file, that is, data stored in a magnetic disk or the like,
The file is provided with data for determining the image representation format of the data, and the data is input to the apparatus together with the image data. As described above, the same applies to image data sent through a communication line such as a facsimile.

This function is effective in a system configuration in which a character recognition system and a plurality of terminals and a plurality of image input devices are connected via a network. This is because there is a possibility that there are various kinds of image input units 1 connected to the network, and these can be freely handled.

Next, the character recognition processing for the image (C) represented in color and the image (G) represented in light and shade will be described.

First, the C or G image is binarized, and the binary image is set as a recognition target. When a suspicious place such as blurring, crushing, or contour distortion occurs during character extraction and character recognition or character recognition, a portion corresponding to the portion of the C or G image is examined. Therefore, in this device, the C or G image is retained even after the binarization.

FIG. 22 is a block diagram. In this case, the input color image is stored as it is or color-separated in the color image buffer 221 as color information. At this time, the color information as it is and the color information after color separation may be stored in separate buffers. Next, a gray image is obtained from the color information by calculating, for example, an average value of RGB density values, and stored in the gray image buffer 222. When the input image is gray, it is input directly into this buffer and the color image buffer is omitted.

Next, a binary image is generated by performing a binarization process on the gray image.
3 is stored. Next, image data for one character is extracted from the data of the binary image buffer 223, and the character is extracted by the character extracting unit 224.
At 25, character recognition is performed. For example, FIGS.
As shown in (c), assuming that patterns A and B are stored as dictionary patterns, X of the input pattern is matched with these patterns. For example, matching is performed by a method such as template matching.

As a result, assuming that the matching degree between A and B is almost the same and a judgment cannot be made, a set of dictionary patterns A and B is provided with a detailed recognition judgment process. It shall be started when judgment is impossible.

In the detailed recognition determination process, a designation is made to check a specific portion of the input pattern.
In this case, there is a designation to check whether a hole is opened in the Y area portion of the input pattern.

Therefore, the density value of the gray image buffer 222 corresponding to this area is checked with reference to the gray image buffer 222. If the density value is closer to the background density value than the surrounding density, it is determined that a hole is opened, and the input character is determined to be "8". In the opposite case, it is determined to be "9".

The same applies when referring to the color image buffer 221 instead of the gray image. Character extraction section 22
4, the gray image buffer 222 and the color image buffer 221 can be similarly referenced.

For example, FIG. 24 shows an example in which two characters are in contact. When the projected component of this image, that is, the number of black pixels, is calculated in the vertical direction, the result is as shown in A of FIG. From the projection component analysis, for example, the minimum point of the projection component, in this case, B and C are set as character extraction candidate points.

Therefore, a cut portion is determined based on these cutout candidate points, and b and c, and the corresponding portions of the gray image or the color image corresponding to those portions are referred to. As a result, when it is determined that the part is close to the background color or close to the density value, it is possible to increase the priority of the cutout candidates of the part. In the case of b and c, it is determined that b has a higher priority as a cut-out point from the value of the projected component, so if it is found that it is close to the density value or the background of the portion of c, c The priority can be raised by setting a portion as a correct cutout point, and as a result, erroneous reading can be prevented.

Of course, the cutout method is not limited to the above method. In addition to this, it is needless to say that the outline information, the width of the character at the time of clipping, and other characteristics may be used, or may be used together.

In general, the above concept is described. When a place that needs to be re-checked determined from the recognition process is determined, the processing of the subsequent stage is fed back, so that the color and shade data of the pattern at that place can be obtained. By referencing, blurring, crushing, contour distortion, etc. are corrected, and line segment extraction, layout understanding, form understanding, detection extraction, character recognition, and the like are performed.

Next, double line cancellation processing in character recognition will be described. The input image is color-separated, and is divided into the colors of the written characters in the portion corresponding to the color of the seal. Then, character recognition is performed on the image data of the entry color. At this time, if two long horizontal lines or one long horizontal line are found in the character description portion, it is determined to be a cancel portion.

A long horizontal line can be detected, for example, by finding a long horizontal line. When the length is not enough, it is configured to check whether they are single characters or contact characters. If such character segmentation or character recognition fails, it is more likely that it is a cancellation.

From this result, it is determined that the character is canceled, and if there is a corrected character in the vicinity, it is recognized as a corrected character. At that time, it is naturally determined that there is a high possibility that the canceled part has been truly canceled. Is done.

Next, a part of the image data of the seal color corresponding to the position of the canceled part is extracted, and it is checked whether or not a sufficient seal color exists in the part. If not, a warning is issued to the operator using the device. Further, the warning is output as a flag, or the presence or absence of the warning is output together with the character recognition result.

For the determination of the seal, a seal determination unit may be provided to determine whether or not the seal exists in the canceled portion based on the shape of the seal color image data. In addition, by introducing a seal stamp determining unit that can determine the similarity of the shape of the seal stamp, it is possible to output which type of seal was found, accompanying the recognition result.

Normally, the color of the seal is different from the entry color. Therefore, if only the color portion of the seal is taken out by color separation and the seal is judged, accurate judgment can be made. The overlapping portion between the entry portion and the seal may be configured to perform the method described in the above-described color separation, or to perform the judgment and collation of the seal based on the possibility that the stamp is pressed on the entry portion.

The case of identifying a character type in character recognition will be described.

In this case, a thick character line indicates that the character has been preprinted, and if such evidence is found in large numbers, the character is cut out for preprinting and the character recognition means is used. Apply Or, it will be applied preferentially.

Finally, the configuration of the correction screen will be described.
From the input image data, a form image of the form is extracted by, for example, color separation processing. The form image may be separated from only the color, or may be extracted by performing the ruled line extraction processing for a specific color. The form image may be extracted in a bit pattern, or may be expressed as vector data, that is, a combination of line segment data including a set of start point coordinates and end point coordinates.

Next, the characters on the form are recognized from the color-separated image data and the like. At this time, information such as the location and size of the recognized character is stored.

Here, the form is determined by a color similar to the color of the form image of the form obtained by the color analysis.
Display an image. At that time, the character of the recognition result is superimposed and displayed at the position where the character existed. At this time, the characters may be displayed in a color corresponding to the entry color.

The similar color may be a color corresponding to a predetermined correspondence, or may be a similar color obtained by calculation.

The whole or a part of the display may be displayed adjacent to the whole or a part of the current image display so that their positional relations are associated with each other.

In the embodiment shown in FIG. 1, the case where the image data supplied from the outside is processed as the input image as it is has been described, but the transmitted image data and the image data stored in the storage device are input. It can also be an image.

In this case, if only the image data is transmitted or stored, there is a disadvantage that the image processing side cannot obtain a clue to perform the color correction.

FIG. 25 shows an example in which the present invention is applied to a communication system. A transmitting section 251 has an image input section 2511, a correction parameter adding section 2512, and a communication line output section 2513, and a receiving section 252 has a communication line input section. 2521, a correction parameter interpretation unit 2522, a color correction unit 2523, and a character recognition unit / image processing unit 2524.

The image read from the scanner is input from the image input unit 2511 as, for example, RGB three-color multivalued color image data. Then, the correction parameter adding unit 2512 corrects the color misregistration such as the code of the scanner, the code representing the serial number, the input device name, the ID number, the model, the modification parameters necessary for the color misregistration correction, or the flag indicating whether the color misregistration has been corrected. Data that can specify the process is given as attribute data to the input image, and is sent from the communication line output unit 2513 to the reception unit 252 having a character recognition function via the communication line 253.

In the receiving section 252, the communication line input section 252
When receiving the data from 1, the modified parameter interpreting unit 25
If the attribute data indicates the model and serial number of the scanner according to 22, the color misregistration correction parameter is interpreted with reference to the correspondence table between the model and serial number and the color misregistration correction parameter registered in advance in the apparatus. At this time, if there is no correction parameter in the correspondence table, error processing is performed, or the color deviation correction parameter is adopted by using a process for automatically obtaining the color deviation correction parameter.

When a correction parameter is directly attached to the received data, this correction parameter is used. If the received data does not include attribute data capable of specifying a correction parameter, error correction processing is performed, or a color registration correction parameter is obtained using a process for automatically obtaining a color registration correction parameter.

If the attribute data has a flag indicating whether or not the color shift has been corrected, and if the flag indicates a color shift,
No color shift correction is performed.

The data thus corrected for color misregistration is sent to the character recognition unit / image processing unit 2524 for processing.

If the system has an image storage device, the attribute data for specifying the above-described color misregistration correction is attached to the input image data and stored, and then the stored data is read out. Thus, the same processing as described above is performed.

In this case, the present apparatus can perform the same processing not only for correcting color misregistration but also for correcting the degree of blur of each color and correcting color bias.

In addition, the present invention is not limited to the above embodiment. In short, the present invention can be variously modified and used without departing from the gist thereof.

[0186]

According to the present invention, when the restrictions on the form color and the entry color, the printing method of the form form and the character part, the character entry method, etc. are loosely applied, and the entry characters are described in the entry column in a random manner. Even at times, characters can be recognized with high accuracy.

[Brief description of the drawings]

FIG. 1 is a diagram showing a schematic configuration of an embodiment of the present invention.

FIG. 2 is a diagram illustrating a state of an RGB signal when a black portion is out of color;

FIG. 3 is a diagram illustrating a state of an RGB signal when a color shift occurs in general.

FIG. 4 is a diagram showing a color misregistration detection mark.

FIG. 5 is a view for explaining a method of correcting color misregistration by matching changing points of RGB signals.

FIG. 6 is a diagram illustrating a state of an RGB signal when a color blur is not uniform.

FIG. 7 is a diagram showing a frequency distribution of density values of a gray image.

FIG. 8 is a diagram showing an example in which a straight line passes through the number “5”.

FIG. 9 is a view for explaining color separation in the example of FIG. 8;

FIG. 10 is a view for explaining color separation in the example of FIG. 8;

FIG. 11 is a view for explaining color separation in the example of FIG. 8;

FIG. 12 is a diagram showing an example in which noise exists near a straight line.

FIG. 13 is a diagram illustrating an example in which a part of a straight line is slightly discolored, and an example in which noise of a color close to the color is attached to the straight line.

FIG. 14 is a diagram showing a state where line segments of different colors intersect.

FIG. 15 is a diagram showing a state where a black line segment and a line segment having any color intersect;

FIG. 16 is a diagram illustrating an example of detecting a partial thickness of a straight line.

FIG. 17 is a diagram showing a method of separating a character and a hatched portion.

FIG. 18 is a view for explaining a process in which the color separation process is limited to only a part.

FIG. 19 is a view for explaining a run-length expression method capable of expressing colors.

FIG. 20 is an exemplary view for explaining a format definition method of run-length data capable of expressing a color.

FIG. 21 is a diagram showing an example in which processing is performed according to the type of an input image.

FIG. 22 is a diagram showing a processing example in the case where blurring, crushing, contour disorder, and the like have occurred.

FIG. 23 is a diagram illustrating a case where it is doubtful whether a part of a character is a loop.

FIG. 24 is a diagram showing an example in which two characters are in contact with each other.

FIG. 25 is a diagram showing an example applied to a communication system.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 ... Image input part 2 ... Color separation part 3 ... Buffer 31, 32, 3n ... Color image buffer 4 ... Image processing part 5 ... Character cutout part 6 ... Character recognition part 7 ... Read result correction part.

 ──────────────────────────────────────────────────続 き Continuing from the front page (72) Inventor Hideo Horiuchi 1st address, Komukai Toshiba-cho, Saiwai-ku, Kawasaki-shi, Kanagawa F-term in the Toshiba R & D Center (reference) 5B029 BB02 CC29 EE08 5B064 AA01 BA01 CA08 CA09 DA03 5L096 AA02 BA17 FA15 FA73 GA38 JA11

Claims (3)

[Claims] 1. A color separation means for performing color separation of a part corresponding to the color of a seal stamp and a color part of an input character with respect to input image data, and an image of a color part of the input character separated by the color separation means. Determining means for performing a cancellation determination on a character description portion of the data; and when the determination means determines that the image data is a canceled portion, extracting a position portion of the image data of the seal stamp color corresponding to the position of the canceled portion A means for outputting the presence or absence of a seal color at the corresponding portion. 2. A character string is recognized by combining color separation means for separating a specific color from input image data and a plurality of methods for recognizing a part separated as a specific color by the color separation means as a character string. An image processing apparatus comprising: 3. A color separating means for separating a specific color from input image data of a form, and a part separated by a color different from the color of the form of the form by the color separating means. An image recognition apparatus comprising: a recognition unit configured to recognize a part which is the same color as the color of the form and is not separated from the form by character recognition.