JP2003115031A - Image processor and its method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To binarize a character boundary area as faithfully as possible and to improve an OCR recognition rate in binarizing process for converting a multi-valued character image into a binary character image. SOLUTION: An image processor is provided with a scanner means for digitizing character information formed on the surface of paper and entering the digital information as multi-valued image information, an arithmetic processing means for executing image processing operation, an image storing means for storing the image information and an image processing result obtained by the arithmetic processing means, a separation means for separating a character boundary area generated by an input from the scanner means, a 1st binarization means for binarizing the character boundary area, a 2nd binarization means for binarizing an area other than the character boundary area, and a correction means for smoothing and correcting the ruggedness of the character boundary area after the binarizing processing.

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 and method, and more particularly to an image processing apparatus and method for performing character image binarization processing for character recognition processing.

[0002]

2. Description of the Related Art With the progress of digitization of documents, the character information formed on the paper surface is often digitized and stored as image information. However, image information in a format such as a bitmap has a large amount of data, and since it is difficult to edit a document (image information), it is desired to convert it into a character code.

Therefore, an OCR (Optical Character Reader) that reads a character image and converts it into a character code is used. When performing the process of converting into a character code by OCR, each character image is processed. Is cut out and the one character image is normalized and recognized based on the direction of the edge of the character boundary area. Therefore, a multi-valued character image digitized by an input device such as a scanner is first
In order to specify the position of the character pixel, the process of separating the character pixel and the background pixel is required.

It is necessary to perform a binarization process using a character image multi-valued by scanner input as an input and output a binary-represented character image. FIG. 27 shows the configuration of a device that performs general character image binarization processing. This device is a scanner unit 1 that digitizes character information formed on paper and reads it as multivalued image data, an image that stores multivalued image data read by the scanner unit 1 and image data after image processing. The storage unit 2 includes a CPU (signal processing unit) 3 that binarizes input multi-valued image data, and an OCR 4 that performs character recognition processing based on the binary-coded character image data. The arrow in FIG. 27 indicates the flow of image information.

The flow of processing is as follows. The multi-valued character image read by the scanner unit 1 is temporarily stored in the image storage unit 2. Next, for the multi-valued character image, C
The PU 3 performs a process of separating the character boundary area generated by the scanner input, and stores the separated image of the character boundary area in the image storage unit 2. Based on the input multi-valued character image stored in the image storage unit 2 and the separated image of the character boundary area created by image processing in the CPU 3, the CPU 3
In step 2, the input multi-valued character image is sequentially subjected to binarization processing and character boundary area smoothing processing. The binary character image data obtained as a result of the binarization processing and the character boundary area smoothing processing is stored again in the image storage unit 2. After saving the binary character image in the image storage unit 2, the OCR 4
Character recognition processing is performed using this binary character image as an input. O
In CR4, since binarization is performed with a simple threshold value, an 8-bit image in which a pixel determined to be a character region has a density value of 255 and a pixel determined to be a background pixel has a density value of 0 is used for character recognition. It is sent to OCR4 as a target image.

As a conventional technique for binarizing a character or a document image, Japanese Patent Laid-Open No. 2000-333022 discloses an image binarizing method and apparatus and a storage medium. Further, Japanese Patent Laid-Open No. 5-282494 discloses an image data binarizing device. Japanese Patent Laid-Open No. 10-308871 discloses a binarizing method and a character reading device. These prior arts are a binarization method and apparatus for a character or a document image, and the density information is used as the information when performing the binarization process.

For this reason, a character image read by an input device, especially a scanner, becomes multi-valued even if the original binary character information is read by the CCD or the MTF characteristic depending on the scanner characteristic, and especially the character having a steep density gradient. The concentration cross section is rounded in the boundary region. The scanner characteristic is unique to each scanner device, and the degree of rounding of the density cross section varies from scanner to scanner. Therefore, in the multivalued character image, the slope of the edge of the density value changes. In particular, in a density cross-section portion of a scanner-input multi-valued image in which the degree of rounding is steep, intermediate density is likely to occur. During the binarization process, it is difficult to determine the binarization of the pixels distributed in the intermediate density area, and in the conventional binarization method, the character image is crushed or blurred due to the erroneous binarization determination of the boundary area. There is a problem of doing.

Further, the degree of rounding of the density cross section is high due to local area information, for example, the area surrounded by a thick character line has a high density value even if the background area is affected by a nearby character area. So it tends to grow. The area that is generated as a result of the density section being blunted by the scanner input and the density section being blunted is mainly the character boundary area, which is the area that transitions from the character pixel to the background pixel or from the background pixel to the character pixel. To do.

Further, particularly, the sampling becomes rough due to the influence of the low resolution, and when digitized, the density value is likely to vary in the character boundary region where the density change is particularly large. FIG. 28 shows how, in the binarization processing of multivalued images having different resolutions, unevenness occurs in the character boundary area in the image after the binarization processing due to the difference in the resolution of the multivalued images. The left side shows an image captured in high resolution by the scanner. The right side shows the image captured at low resolution by the scanner. In particular, in the low-resolution image in the upper right diagram, since the sampling interval is large, it is not possible to express a steep density gradient in the character boundary area, and density variations occur. The result of binarizing these images with a simple threshold (density value 128, * 1 pixel 8 bit expression) is shown in the lower diagram. As described above, particularly in a low-resolution image, since a steep density gradient in the character boundary region cannot be expressed, unevenness is likely to occur in the image binarized by the conventional binarization method.

When the multi-valued character image input by the scanner is binarized by using only the density information, as described above, the degree of rounding of the density cross section caused by the scanner characteristic is locally different. As a result, the characters are crushed or blurred and the OCR recognition rate is lowered. In the conventional binarization method, the density section generated by the scanner input is blunted, and the binarization determination is often erroneous in the region resulting from the blunted density section. There is a demand for a binarization processing method and apparatus that are less likely to be erroneously determined for a region resulting from a blunted density cross section in order to improve the OCR recognition rate.

Further, as described above, there is a problem that unevenness is likely to appear in the character boundary area due to low resolution input, and the OCR recognition rate is lowered. The OCR recognizes by detecting the direction of the character boundary in the boundary area of the character image. O
In order to improve the CR recognition rate, a feature amount for recognition is used. In the conventional method, there are many erroneous determinations in the binarization processing.
There is a need for a method and apparatus that can faithfully binarize information in character boundary areas.

[0012]

SUMMARY OF THE INVENTION The present invention has been made in view of the above problems, and it is a binarization for outputting a binary character image from a multivalued character image for character recognition processing in OCR. It is an object of the present invention to provide an image processing method and apparatus capable of binarizing a character boundary area in processing as faithfully as possible and improving the OCR recognition rate.

According to the first aspect of the present invention, for the character boundary area generated by the scanner input, the information of the density cross section, which is the local information of the character, is used for the binarization determination, so that the character boundary area is binarized. The purpose is to suppress erroneous determination and improve the OCR recognition rate.

Since a character image has a high contrast, the character boundary area of the character particularly in the scanner-inputted character image is deteriorated and liable to be blunted. The character boundary area generated by the scanner input is especially affected by the local position information of the character lines and the font thickness, as the degree of deterioration of the character boundary area is affected by the density information of the neighboring areas. In the conventional method in which a single threshold value of density value is set for a pixel, a character is crushed or blurred after binarization. The invention according to claim 2 prevents the erroneous binarization of the boundary region of the character.
Extract the boundary area of characters generated by scanner input,
The purpose is to perform binarization suitable for this region and improve the recognition rate of OCR.

Since the character image has a high contrast, the character input area of the scanner is apt to be blunted due to deterioration of the character boundary area. Since the boundary area of characters generated by scanner input also changes depending on the local position information of character lines and the thickness of fonts, especially in the conventional method of setting a single density value threshold value for all pixels. The characters were crushed or blurred after the value was converted. According to the third aspect of the present invention, by using the LBG algorithm for pixels that are difficult to determine the binarization, the character boundary area generated by this scanner input is extracted by performing efficient separation, and is suitable for this area. The purpose is to perform binarization processing and improve the recognition rate of OCR.

In the character boundary area generated by scanner input, character pixels and background pixels are mixed. In order to perform the binarization process on the region, a local binarization threshold value is provided for each concentration cross section, and the binarization process is performed for each concentration cross section to perform the local binarization process. The object of the present invention is to prevent crushing and blurring of a character image after binarization by setting a threshold value for thresholding, and to improve the recognition rate of OCR.

According to a fifth aspect of the present invention, for an image in which the character boundary region generated by scanner input is binarized, the region inside the character and the background region are stably used by using the neighboring binarization results. It is intended to be binarized.

When character recognition is desired by OCR, in particular,
Due to the influence of the low resolution, erroneous recognition may occur due to unevenness generated in the character boundary area after the binarization process. It is an object of the present invention to improve the OCR recognition rate by removing unevenness generated in the boundary area by smoothing the character boundary area in the binarized character image.

The character boundary area generated by the scanner input is apt to be blunted due to deterioration of the character boundary area due to the influence of the scanner characteristics and the character font. Since a character image read by a scanner or the like is also affected by the character vicinity area, the density value greatly differs between the character area and the background area. When a single density threshold value is set for all pixels in an image input by a scanner and the image is binarized, the characters are crushed or blurred. According to a seventh aspect of the present invention, the detection unit separates the character boundary region generated by the scanner input and switches the binarization process for each region, thereby making an error in binarizing the character boundary electronic region after binarization. The object is to suppress the judgment and improve the recognition rate of OCR.

In the character boundary area generated by scanner input, character pixels and background pixels are mixed. In order to binarize the region, a local binarization threshold value is provided for each cross section and binarization processing is performed for each density cross section. By locally setting a binarization threshold value using the shape of the density cross section, it is an object to prevent the character image after binarization from being crushed or blurred and to improve the OCR recognition rate.

[0021]

In order to achieve such an object, the invention according to claim 1 is a scanner means for digitizing character information formed on a paper surface and taking it in as multi-valued image information, and image processing. Arithmetic processing means for performing arithmetic,
An image processing apparatus comprising: an image storage unit for storing image information and an image processing result by an arithmetic processing unit; a separating unit for separating a character boundary region generated by an input by a scanner unit; First to process
The binarizing means, the second binarizing means for binarizing the area other than the character boundary area, and the correcting means for smoothing and correcting the unevenness of the character boundary area after the binarizing processing. Is characterized by.

The invention according to claim 2 is characterized in that, in the invention according to claim 1, the separating means separates the character boundary area by using the information of the primary differential value and the density value as the characteristic amount. .

According to a third aspect of the invention, in the first aspect of the invention, the separating means separates the character boundary area by the LBG algorithm using the first differential-density plane of the input image. .

According to a fourth aspect of the present invention, in the first aspect of the present invention, the first binarizing means binarizes the character boundary area using the character density cross-section information generated by the input by the scanner means. It is characterized by doing.

According to a fifth aspect of the present invention, in the first aspect of the present invention, the second binarizing means performs the binarizing process by using the binarizing result of the neighboring character boundary area. It has a feature.

According to a sixth aspect of the present invention, in the first aspect of the invention, the correction means performs smoothing by detecting and filling irregularities generated in the character boundary area of the image after the binarization processing. It has a feature.

The invention according to claim 7 is characterized in that, in the invention according to claim 1, the character boundary area is separated, and the binarization processing is switched for each separated area.

According to an eighth aspect of the present invention, in the first aspect of the present invention, the first binarizing means uses a character boundary area as a character density cross section produced by an input by the scanner means and information on its shape. It is characterized by performing binarization processing.

According to a ninth aspect of the present invention, with respect to a multi-valued character image input by a scanner, a separation step of separating a character boundary area generated by the scanner input, and a first binary processing of the separated character boundary area. It has a binarization step, a second binarization step for binarizing an area other than the character boundary area, and a correction step for smoothing and correcting irregularities in the character boundary area after the binarization processing. I am trying.

The invention according to claim 10 is characterized in that, in the invention according to claim 9, the separating step separates the character boundary area by using the information of the primary differential value and the density value as the feature amount. .

In the invention described in claim 11, in the invention described in claim 9, the separating step is the first derivative of the input image.
It is characterized in that the character boundary area is separated by the LBG algorithm using the density plane.

According to a twelfth aspect of the present invention, in the ninth aspect of the present invention, the first binarizing step binarizes the character boundary area using the character density cross section information generated by scanner input. Is characterized by.

According to a thirteenth aspect of the present invention, in the ninth aspect of the invention, the second binarization step performs the binarization process by using the binarization process result of the adjacent character boundary region. It has a feature.

According to a fourteenth aspect of the invention, in the ninth aspect of the invention, the correction step performs the correction by detecting and filling the unevenness generated in the character boundary area of the image after the binarization processing. I am trying.

The invention of claim 15 is characterized in that, in the invention of claim 9, the character boundary region is separated and the binarization processing is switched for each separated region.

According to a sixteenth aspect of the invention, in the ninth aspect of the invention, the first binarizing step uses the character boundary area by using the character density cross section generated by the scanner input and the information of the shape. The feature is that it is digitized.

[0037]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[First Embodiment] FIG. 1 is a diagram showing an image processing method according to the first embodiment of the present invention. This image processing is processing performed by the CPU (signal processing unit).
In the image processing method, the input multi-valued character image I1
A character boundary region extraction process 101, a multi-valued character image I1 and a character boundary region binarization process 102 based on a character boundary region separation image I2 obtained after the character boundary region extraction process, and a character Binarization processing 103 of the area other than the character boundary area is performed on the image after the binarization processing of the boundary area, and correction processing 104 of the character boundary area is performed on the output image after the binarization processing. I3 is output.

In the character boundary area extraction processing 101, the character boundary area generated by the scanner input is separated based on the characteristic amount such as the first derivative (of the density curve).

In the binarization processing 102 of the character boundary area, attention is paid to the boundary area of the character generated by the scanner input, and the threshold value for determination is locally changed to perform the binarization processing. Figure 1
FIG. 9 shows an example of Mincho typeface and Gothic type character images captured by the scanner and an example of the binarization result. In FIG. 19, the upper image shows the original image. The upper and lower figures in the center show the density distribution on the density cross section of the original image. The lower image shows an image as a result of binarization processing with each of the threshold A and the threshold B shown on the density distribution. In this process, as shown in FIG. 19, raster scanning is performed in the horizontal direction of the target character image, and the density section is set for each horizontal direction. The edge portion, which is the character boundary area, has the above-described character boundary area extraction processing 10
The resulting image separated by 1 is used.

The threshold value A in the center diagram is a simple threshold value of 128, which is an intermediate density value of an 8-bit image. On the other hand, the threshold value B is set to an intermediate value between the maximum density pixel and the minimum density pixel in the area (dotted line part in the upper part of the central figure) determined to be the edge area which is the character boundary area in the character boundary area extraction processing 101. It has been set. The threshold value B is similarly set for the other edge portions which are the character boundary areas. The lower image in FIG. 19 is the binarization processing result after the above processing is fully scanned in the horizontal direction. As can be seen from the lower image of FIG. 19, when the simple threshold A is set from the entire image and binarization is performed, the character boundary area generated by scanner input is blurred in the Mincho type with thin character lines. Occurrence occurs in the Gothic type with thick character lines. On the other hand, in the method in which the local threshold value B is set for each one density cross section (edge portion), the peaks and valleys of all density cross sections can be reproduced, so that blurring or crushing of characters is less likely to occur.

In the binarization process 102, only the edge portion which is the character boundary area is binarized in practice.
For comparison with the conventional method using the simple threshold value, an example in which pixels other than the character boundary area are binarized with the threshold value A is shown.

Binarization processing of the area other than the character boundary area 10
In No. 3, since the character boundary area generated by the scanner input in the process of 102 is binarized, the character boundary area generated by the binarized scanner input in the vicinity of the inner area of the character line and the background area is binarized. By using and integrating the information of, stable binarization processing is performed.

In the character boundary region correction processing 104, sampling is rough due to the influence of low resolution, and especially in the character boundary region where the density change is large when digitized, the character boundary region in which the density value variation easily occurs. A correction process is performed for it. Since it is not possible to express a steep density gradient in the character boundary area in a low-resolution image, unevenness is likely to occur in the binarized image. It is removed by the smoothing process of.

FIG. 2 is a diagram showing the flow of binarization processing based on the basic configuration of the image processing method of the present invention shown in FIG. In contrast to the character boundary area extraction process 101 in FIG. 1, in the process of FIG. 2, a non-determined pixel separation process 201 is performed. With respect to the multi-valued character image captured by the scanner unit as the input image, pixels distributed in the character boundary area due to scanner input due to a sudden change in density in the character boundary area are defined as non-determined pixels P1, and the character boundary area is defined. In other cases, the separation processing is performed by setting the pixels distributed in the area where the density change in the vicinity is flat as the definite pixel P2.

FIG. 12 shows a diagram of an image which has been subjected to the non-determined pixel separation processing 201. The diagram on the left is an example of a multi-valued character image captured by scanner input, and the image on the right is the image obtained by subjecting the original image to the non-determined pixel separation processing 201. The non-determined pixel P1 is shown in white and the determined pixel P2 is shown in black. As shown in the right figure, the non-deterministic pixels P1 are the character boundary area generated by the scanner input, and therefore are distributed in the character blurred area generated by the scanner characteristics. Since the fixed pixels P2 are areas other than the character boundary area generated by the scanner input, they are distributed in the area inside the character line and the background area that is not affected by the character. Pixels mainly distributed in the character boundary area are defined as non-determined pixels P1. Pixels distributed in the background area where there is no influence of characters and the area inside the character line are defined as fixed pixels P2. As a specific separation method, the non-determined pixel P1 performs the separation process using the first-order differential characteristics of the density as a feature amount by utilizing the fact that the density change in the vicinity is rapid. An edge detection filter is applied to the input multivalued character image to calculate the strength of the first derivative. A threshold value is set for the intensity of the first-order differential, and a pixel above the threshold value is set as the non-determined pixel P1 and a pixel below the threshold value is determined pixel P2 to obtain the separated image 208 of the character boundary region.

Next, as the binarization processing 102 of the character boundary area, the binarization processing 202 of the undetermined pixel P1 is performed. The input is the multi-valued character image captured by the scanner unit and the separated image of the character boundary area. The multi-valued character image is used to determine the threshold value from the density information. In the separated image of the character boundary area, the target image is an undetermined pixel P distributed in the character boundary area.
It is used to check if it is 1. Binaryization is performed by setting a local square neighborhood area centered on the non-determined pixel of interest and using the average density value of the pixels distributed in that neighborhood as the threshold of the non-determined pixel of interest. Perform processing.

Next, as the binarization processing 103 of the area other than the character boundary area, the binarization processing 203 of the fixed pixel P2 is performed. As an input, an image in which only the character boundary region generated by the scanner input is binarized after the binarization process 202 of the undetermined pixel P1 is used. In order to binarize the peripheral area of the character boundary area in which the non-determined pixel P1 is distributed, the characters near the determined pixel P2 are determined based on the information of the non-determined pixel P1 distributed in the vicinity of the determined pixel of interest. The number of pixels or the number of background pixels is calculated to make a determination, and the pixel with the larger number is integrated.

Next, the character boundary area correction processing 204 is performed with respect to the character boundary area correction processing 104 of the image processing method according to the first embodiment. As an input, an image obtained by binarizing all pixels in the binarization processing 203 of the fixed pixel P2 is used. X direction (horizontal) and Y direction (vertical) as shown in FIG.
Raster scanning is performed from the upper left pixel of the to detect a continuous area of horizontal and vertical character pixels, and if the continuous area is narrow, fill that pixel with a character pixel to create a horizontal and vertical character boundary. Fill the irregularities. In FIG. 13, in the examples of 2 and 4, white pixels (background pixels) that occur between consecutive black pixels (character pixels) are corrected to black pixels (character pixels).

Next, a continuous region of horizontal and vertical background pixels is detected from the upper left pixel in the X and Y directions as shown in FIG. 13, and when the continuous region is narrow, the pixel is set as the background. By filling in the pixels, the unevenness of the horizontal and vertical character boundaries is filled. In FIG. 13, in the examples of 1 and 3, the black pixels (character pixels) generated between the continuous white pixels (background pixels) are corrected to white pixels (background pixels). The binary character image in which the character boundary area is corrected by filling the irregularities in the boundary area of the character image is output.

FIG. 3 shows a flow of binarization processing which is a variation different from the image processing method of FIG. The difference from the method of FIG. 2 is that in the binarization process 303 of the fixed pixel P2, the input multi-valued character image I1 and the non-fixed pixel P1 are input.
The binarization-processed image is used to binarize the determined pixel P2. Since the fixed pixel P2 can be separated into a sufficiently high density value and a sufficiently low density value, a threshold value is set for the density value to binarize the fixed pixel P2 without using nearby information as in the method of FIG. Other processes are the same as the method of FIG.

[Second Embodiment] FIG. 4 shows the undefined pixel P in the image processing method according to the second embodiment of the present invention, in contrast to the character boundary area extraction processing 101 described in the first embodiment.
The separation processing of No. 1 is shown. In the separation process of the non-determined pixel P1, the region is assumed from the density cross section of the character image, the multi-valued character image I1 is input, and the non-determined pixel P1 distributed in the character boundary region generated by the scanner input is extracted.

FIG. 14 shows a density cross section of a character image. Area 1 is a background area with a flat density gradient, area 2 is a transition area from a character to a background / background to a character, area 3 is a character area with a flat density gradient, and area 4 is a character line. And the transition region to the narrow background region sandwiched between the character lines. First, assuming that there is a region as shown in FIG. 14, regions 2 and 4 in which the undetermined pixel P1 is distributed are extracted based on the intensity of the first derivative and the density information. The intensity of the first derivative is obtained by performing the calculation process 401 of the first derivative.

FIG. 15 shows the first derivative and the distribution of density values.
Regions 1 to 4 in FIG. 14 are distributed as shown in FIG. 15 when viewed on the first derivative-density value plane. On this plane,
The undetermined pixels P1 distributed in the regions 2 and 4 and the undetermined pixels P2 distributed in the regions 1 and 3 are separated using a threshold value T1 as shown in the figure. Separation processing is performed with the area above the threshold value T1 as the area of the non-determined pixel P1 and the area below it as the area of the determined pixel P2.

Threshold T on the first derivative-density value plane of FIG.
In the setting processing 402 of No. 1, the input multi-valued character image I1
A line that is linearly divided from the reference point with the average value of the maximum and minimum density values in the horizontal direction as the reference and the coordinate in the vertical direction as the coordinate 0 is the threshold T1. When the threshold value T1 is determined, the distribution of pixels existing in a straight line is searched from the reference point, two minimum straight lines are obtained, and the straight lines are set as the threshold value T1.
After the threshold value T1 is obtained, by the separation processing 403 at the threshold value T1, the area above the threshold value T1 on this first-order differential-density value plane is set as the area of the undetermined pixel P1 from the input multivalued character image I1. Extract. Then, the separated image I2 of the character boundary area is obtained.

[Third Embodiment] FIG. 5 shows a third embodiment of the character boundary region extraction processing 101 described in the first embodiment.
7 shows a flow of separation processing of undetermined pixels in the image processing method in the embodiment. In the separation process of the non-determined pixel P1, the region is assumed from the density cross section of the character image, and the non-determined pixel P1 distributed in the character boundary region generated by the scanner input is extracted. Regions 1 to 4 shown in FIG. 3 of the second embodiment are efficiently separated using the LBG algorithm on the first derivative-density value plane shown in FIG.

In the non-determined pixel separation processing in the third embodiment, the input image is the same scanner-input multi-valued character image I1 as in the second embodiment, and the maximum and minimum first-order differential values,
Extract the maximum and minimum density values. The intensity of the primary differential is obtained by performing the same process as the calculation process 401 of the primary differential value of the second embodiment.

Next, the LBG on the first derivative-density value plane
Perform grouping process 502 by algorithm,
The non-determined pixel P1 is separated. In order to divide the area shown in FIG. 15 into five areas, the LBG that is the center of the five areas
The initial point of the algorithm is set as shown in FIG. Next, 5 is applied to all distribution points by the LBG algorithm.
The minimum distance between points is calculated, the distance to which central point is closest is calculated, and the points are grouped into the closest points. FIG. 17 shows a distance calculation formula between each center point and each distribution point. A graph in which the coordinates of the distribution point are (x, y) and the coordinates of the center point of each region are (x ₀ , y ₀ ) with the density value in the x direction and the primary differential value in the y direction on the first derivative-density value plane. The distance between the distribution point and the center point is calculated by the lower calculation formula 21. The center point of the group is calculated from the grouped points, distance calculation is performed to find the closest point again, the closest point is searched, and the point is grouped again. The calculation calculation of the center point is repeated, and the process ends when the movement distance of the center point converges. At that time, the pixels grouped at the center points of the regions 1 and 3 are separated as definite pixels P2, and the pixels grouped at the center points of the regions 2 and 4 are separated as non-determined pixels P1. Then, the separated image I2 of the character boundary area is obtained.

[Fourth Embodiment] FIG. 6 shows an image processing method according to a fourth embodiment of the present invention in contrast to the binarization processing 102 of the character boundary area described in the configuration of the first embodiment. 2 shows a flow of binarization processing of non-determined pixels of. In this undetermined pixel binarization process, the pixels in the character boundary region are locally binarized based on the information of the density cross section.

For the input of the binarization processing in the fourth embodiment, the multi-valued character image I1 input by the scanner and the separated image I2 of the character boundary area are used.

FIG. 18 shows an example of the density cross section of a character image. Targeting the undetermined pixels P1 distributed in the character boundary area generated by the scanner input, the threshold value T2 is set as shown in the figure for each gradient of the cross section, and the density of the undetermined pixels P1 belonging to the one density cross section is set. Binarization processing is performed using local linear information called a cross section.

FIG. 19 shows a region near the undefined pixel P1. Since the character image is a two-dimensional image, there are density cross sections in four directions. Binarization processing is performed using the information on the character boundary area generated by the scanner input in these four directions. First, the multi-valued character image I1 is subjected to the primary differential calculation processing 601 by the directional differential operator of FIG. Raster scanning is performed in the horizontal direction with respect to the image, and the undetermined pixel P of interest
Is searched for, and a primary differential value (inclination of the density cross section) is calculated by applying a filter weighted for each direction around the pixel P. This is calculated as a feature amount for investigating which direction of the binarization result shown later is more effective. As the inclination of the density cross section becomes larger, it is clear where the non-determined pixels P are distributed in the character boundary region, and it is considered to be effective for the binarization determination.

After the primary differential calculation processing 601, a binarization result acquisition processing 602 in the vicinity of the undetermined pixel P1 in the + 45 ° direction is performed. The non-determined pixels P1 existing in the direction from the target pixel P to the determined pixel P2 and distributed in that direction are set as the neighboring regions, and the local threshold value T2 is set for each of the neighboring regions. FIG. 20 shows the density distribution of the undetermined pixel P1 which is determined to be the neighboring area distributed in the + 45 ° direction of FIG. 1
The threshold value T2 of the neighborhood area is set for each direction and binarization processing is performed.

Thereafter, the undetermined pixel P1 in the + 45 ° direction
Similar to the binarization result acquisition process 602 in the vicinity region of, the direction is changed and the same process is performed. Undetermined pixel P in the horizontal direction
Acquisition processing 603, -45 of the binarization result in the neighborhood region of 1
A binarization result acquisition process 604 in the vicinity region of the non-determined pixel P1 in the ° direction and a binarization result acquisition process 605 in the neighborhood region of the non-determined pixel P1 in the vertical direction are performed in order. Obtain the binarization result of. The final binarization result is obtained by determining the inclination of the concentration cross section, which is the above-mentioned first-order differential value, in the binarization result for each direction by the voting process 606. The above process is performed on all the undetermined pixels P1 and the selected pixels. This voting process is, for example, a horizontal binarization result is a character with a primary differential value of 50, a vertical binarization result is a background with a primary differential value of 100, and a + 45 ° binary conversion result is a background. Then, the primary differential value is 10, and the binarization result in the −45 ° direction is a character, and the primary differential value is 35. In this case, the first differential value of the horizontal direction and the −45 ° direction is voted to 85 for the character, and the first differential value of the vertical direction and the + 45 ° direction is voted to 110 for the background, and as a result, the final differential value is obtained. The pixel binarization result becomes the background.

All the pixels determined to be the non-determined pixel P1 are subjected to the raster scanning in the horizontal direction and searched, and the above processing is performed to obtain the binarized image I4 of only the region of the non-determined pixel.

[Fifth Embodiment] FIG. 7 shows the image processing method according to the fifth embodiment of the present invention, which is different from the binarization processing 103 other than the character boundary area described in the first embodiment. 7 shows a flow of binarization processing of fixed pixels. In the binarization processing of the confirmed pixels, the confirmed pixels P2 distributed in the area inside the character and the background area are binarized by using the information of the adjacent unconfirmed pixels P1 that are binarized.

The input image is the binarized image I4 of only the region of the non-determined pixel P1 in which the character boundary region generated by the scanner input is binarized in the non-determined pixel binarization process 202 of the first embodiment. To use.

Determining process 701 of the area near the determined pixel P2
The area used for binarization of the fixed pixel P2 is determined by. FIG. 21 shows a region near the fixed pixel P2. In this way, 8 neighboring pixels of the focused pixel P2 of interest are searched and determined as a neighboring region.

After the neighborhood area is determined, the voting process 702 is performed to perform the binarization process of the fixed pixel P2. The number of binarized undetermined pixels P1 distributed in the location determined as the neighboring region is checked and voted, and binarization is performed by integrating the voted majority. In the case where the numbers are the same, the enlargement processing 703 of the neighborhood area of the fixed pixel P2 is further performed, and the neighborhood area is enlarged as shown in the right diagram of FIG. 702 is performed to perform binarization. The raster scanning is performed in the horizontal direction from the upper left of the target image to search for the fixed pixel P2, and the above processing is performed. Definite pixel P2 being noticed
When only the confirmed pixel P2 exists in the neighborhood area of, the binarization result of the already processed confirmed pixel P2 is also reflected in the vote and binarized.

FIG. 22 shows the result of the above-described processing for an image in which the character boundary area is binarized and the fixed pixel P2 is not binarized. The fixed pixel P2 surrounded by the background as shown in FIG. 22 is binarized into the background pixel, and the fixed pixel P2 surrounded by the character pixels is binarized into the character pixel.

All the pixels determined to be the definite pixel P1 are subjected to the definite pixel binarization process described above to obtain a binary character image (before character boundary area correction) I5.

[Sixth Embodiment] FIG. 8 shows a character boundary area in the image processing method according to the sixth embodiment of the present invention, in contrast to the character boundary area correction processing 104 described in the first embodiment. The flow of the correction process of is shown. In the correction process for the character boundary region, the unevenness generated in the character boundary region is removed by performing a process of filling it by smoothing.

The input image is the binarization processing 2 of the determined pixel P2.
A binary character image (before character boundary area correction) I5 in which all pixels after 03 are binarized is used.

As shown in FIG. 13, for the character image,
Raster scanning is performed twice in the x (horizontal) direction to fill in the background pixels distributed between the continuous regions of the character image in the horizontal direction 80.
1 and the padding process 802 of the character pixels distributed between the continuous regions of the background pixels in the horizontal direction. In the first x-direction raster scan, character pixels are searched for, and when a continuous area of character pixels is extracted, it is searched whether background pixels continue in that direction. If it continues, the continuous area of the next character pixel is searched up to the last pixel in the x direction. If not followed,
Since the character image is distributed again after the background pixel, the background pixel is embedded in the character pixel to perform a process of connecting continuous areas of the character pixel. This process is performed for all x directions. next,
Focusing on the background pixels, the same processing as described above is performed on the continuous area of the background pixels, and the processing of connecting the continuous areas of the background pixels is performed in all x directions.

Similarly in the y (vertical) direction, raster scanning is performed twice to fill in the gaps 803 between the continuous regions of the vertical character pixels and the characters distributed between the continuous regions of the vertical background pixels. A pixel filling process 804 is performed.

FIG. 23 shows an example of smoothing processing of the character boundary area. 23 (a) and 23 (c), the unevenness of the boundary area, which is considered to be noise and is present in the boundary area of the character, is removed by smoothing to remove the unevenness of the character as shown in FIGS. 23 (b) and 23 (d). The binary character image I3 whose boundary area has been corrected is output. Focusing on the boundary of the binarized character image, the unevenness of the boundary area generated during the binarization process due to the influence of the low resolution is extracted, smoothed, and removed. A continuous area of character pixels in the vertical direction is detected, and when the interval between the continuous areas is equal to or smaller than a threshold value, the pixel is filled with the character pixel to fill the unevenness of the character boundary area. Similar processing is performed on a continuous area of character pixels in the horizontal direction, and when the interval between the continuous areas is less than or equal to a threshold value, the pixels are filled in with the character pixels to fill the unevenness of the character boundary area. Further, the same processing is performed for the continuous area in the vertical and horizontal directions of the background pixel to fill the unevenness of the boundary area of the character image. FIG. 24 shows an execution example of the smoothing process.

[Seventh Embodiment] FIG. 9 shows a process for switching the binarization process for each region in the image processing method according to the seventh embodiment of the present invention. In the separation processing of the non-determined pixel P1, the region is assumed from the density cross section of the character image, and the non-determined pixel P distributed in the character boundary region generated by the scanner input.
Extract 1. First, in the selection processing 901, the regions are separated by using the feature amounts such as the primary differential and the density value. Then, different binarization processing 902, 903 is performed for each area.

FIG. 12 shows a density distribution chart of an image which has been subjected to the non-determined pixel separation processing 201 of the second embodiment. The non-determined pixel P1 which is the character boundary area is extracted from the first-order derivative and the density information, and the local threshold value T2 in the method of the seventh embodiment is set for the pixel that is largely affected by the scanner input. Set and binarize.

Further, FIG. 25 shows a distribution diagram of the non-determined pixel P1 and the determined pixel P2 in the density cross section. From this figure, it can be seen that the non-determined pixels P1 are distributed in the character boundary area corresponding to the character boundary area, and the determined pixels P2 are distributed in the area where the density change is flat. By separating the pixels distributed in the character boundary area generated by the scanner input from the pixels distributed in areas other than the character boundary area, 1 is set for the non-determined pixel P1 distributed in the character boundary area generated by the scanner input.
A threshold value T2 is set for each concentration cross section, and binarization processing is performed.

After the binarization processing is performed by each of the binarization methods, the integration processing 904 of the results integrates the binarized results separated in the selection processing 901, and the results are finally binarized. The value character image I3 is output.

[Eighth Embodiment] FIG. 10 shows the binarization processing 102 of the character boundary area described in the first embodiment.
13 shows a flow of binarization processing of undetermined pixels in the image processing method according to the eighth embodiment of the present invention. In the binarization processing of the non-determined pixels, the local binarization processing is performed on the pixels in the character boundary area based on the information of the density cross section. The input of this processing uses the multi-valued character image I1 by the scanner input and the separated image I2 of the character boundary area. The fourth embodiment described above is different from the binarization result acquisition processing 1002 in the neighborhood area of the non-determined pixel P1 in the + 45 ° direction by the new threshold in the neighborhood area of the non-determined pixel P1 in the vertical direction. The threshold value setting method in the processes up to the binarization result acquisition process 1005 is different.

In the binarization result acquisition processing 1002 with a new threshold value in the neighborhood area of the non-determined pixel in the + 45 ° direction, the non-determined pixel P1 existing in the neighborhood area up to the position of the confirmed pixel P2 and distributed in that direction. Is set as a neighborhood area, and a local threshold value is set for each neighborhood area. In FIG. 26,
A threshold shift for preventing blurring and crushing of characters is shown. Fourth
In the embodiment of FIG.
Of the maximum and minimum densities is set as the threshold value T2.
Is set to. On the other hand, in the eighth embodiment, the threshold value T2 is shifted using the shape of the density distribution of the non-determined pixel P1 in which the pixel of interest is distributed so that the small peaks and valleys of the density section can be reproduced. , Prevent blurring and crushing of characters. As shown in (a) of FIG. 26, when the concentration cross-sections are distributed so as to draw mountains, the threshold value is lowered to prevent blurring of characters. When the shape of the density cross section is distributed so as to draw a valley as in (b), the threshold value is increased to prevent the characters from being crushed. For other shapes, or for monotonous increase or decrease, the threshold shift is not performed.

In the voting process 1006, the same process as in the fourth embodiment is performed to obtain the final binarization result. The non-determined pixel P1 is searched in the horizontal direction of the image, the above-described processing is performed on all the non-determined pixels P1, and the binarized image I4 only in the region of the non-determined pixel P1 is obtained.

The embodiments of the present invention have been described above. The above-described embodiment shows an example of a preferred embodiment of the present invention, and the present invention is not limited thereto, and various modifications can be made without departing from the scope of the invention. .

[0085]

As is apparent from the above description, according to the first aspect of the invention, the conventional binarization method extracts the character boundary area caused by the scanner input having a high probability of erroneous binarization determination. Since the binarization target pixels are distributed in the character image based on the information of the density cross section, the binarization is performed, so that the image quality of the character image after output can be improved. In addition, the OCR recognition rate can be improved.

According to the second aspect of the invention, the character boundary area generated by the scanner input is separated from the character image. With the conventional binarization method, it is possible to change the binarization process for a character boundary region generated by a scanner input that has a high probability of making an error in the binarization determination.

According to the third aspect of the invention, the character boundary area generated by the scanner input from the character image is
Separation is efficiently performed by using the LBG algorithm. The LBG algorithm can divide the region by the number of initially set initial points. With the conventional binarization method, it is possible to change the binarization process for a character boundary region generated by a scanner input that has a high probability of making an error in the binarization determination.

According to the fourth aspect of the invention, it is possible to set a local threshold value for the character boundary area generated by the scanner input by using the information of one density cross section. Since the threshold value is set from the 1-density cross section, it is possible to extract the background area surrounded by thick character lines and faithfully reproduce thin character lines. Further, the OCR recognition rate is improved.

According to the fifth aspect of the present invention, in the area other than the character boundary area generated by the scanner input, it is erroneously determined as the background inside the character, and the character area is less likely to exist in the background area. The image quality of the output character image is improved, and the recognition rate is improved in the OCR sensitive to noise of the granular character image in the character area.

According to the sixth aspect of the invention, since the unevenness of the character boundary area generated after the binarization processing is removed by smoothing by the hole filling processing, the image quality of the character image is improved and the character image is improved. OC sensitive to irregularities in the boundary area
In R, the recognition rate is improved.

According to the seventh aspect of the invention, it is possible to perform a local binarization process in which the local threshold value is changed by using the information of one density cross section for each character boundary region generated by the scanner input. Becomes Since the threshold value is set from the 1-density cross section, it is possible to extract the background area surrounded by thick character lines and faithfully reproduce thin character lines. Also, O
CR recognition rate is improved.

According to the eighth aspect of the present invention, it is possible to set a local threshold value for a character boundary area generated by scanner input using information on a density cross section of a character and its shape. Since the threshold is changed and set based on the information of the character density cross section and its shape, it is possible to extract the background area surrounded by the thick character line and faithfully reproduce the thin character line. Further, the OCR recognition rate is improved.

[Brief description of drawings]

FIG. 1 is a diagram showing an image processing method according to a first embodiment of the present invention.

FIG. 2 is a diagram showing the flow of binarization processing of the image processing method according to the first embodiment of the present invention.

FIG. 3 is a diagram showing another method of the binarization processing of the image processing method according to the first embodiment of the present invention.

FIG. 4 is a diagram showing an undetermined pixel separation process of an image processing method according to a second embodiment of the present invention.

FIG. 5 is a diagram showing an undetermined pixel separation process of an image processing method according to a third embodiment of the present invention.

FIG. 6 is a diagram showing binarization processing of an image processing method according to a fourth embodiment of the present invention.

FIG. 7 is a diagram showing binarization processing of fixed pixels in the image processing method according to the fifth embodiment of the present invention.

FIG. 8 is a diagram showing a correction process of a character boundary area in the image processing method according to the sixth embodiment of the present invention.

FIG. 9 is a diagram showing a binarization process switching process of an image processing method according to a seventh embodiment of the present invention.

FIG. 10 is a diagram showing non-determined pixel binarization processing of an image processing method according to an eighth embodiment of the present invention.

FIG. 11 is a diagram showing an example of binarization processing using a simple threshold value A and a local threshold value B.

FIG. 12 is a diagram showing an image on which non-determined pixel separation processing has been performed.

FIG. 13 is a diagram showing removal of irregularities in a character boundary region.

FIG. 14 is a diagram illustrating a density cross section of a character image.

FIG. 15 is a diagram showing the distribution of primary differentials and density values.

FIG. 16 is a diagram showing a separation process of an undetermined pixel P1 by the LBG algorithm.

FIG. 17 is a diagram showing a distance calculation formula between a center point and a distribution point.

FIG. 18 is a diagram showing an example of a density cross section of a character image.

FIG. 19 is a diagram showing a region near a non-determined pixel P1.

FIG. 20 is a diagram showing a density distribution of an undetermined pixel P1 which is determined as a neighboring area distributed in a + 45 ° direction.

FIG. 21 is a diagram showing a region near a fixed pixel P2.

FIG. 22 is a diagram illustrating an example of execution of binarization processing of a fixed pixel P2.

FIG. 23 is a diagram illustrating a smoothing process for a character boundary area.

FIG. 24 is a diagram illustrating an execution example of smoothing processing of a character boundary area.

FIG. 25 is a non-determined pixel P1 and a determined pixel P2 in a density section.
It is a figure which shows about distribution of.

FIG. 26 is a diagram showing a shift of a threshold value for preventing blurring and crushing of characters.

FIG. 27 is a diagram showing a configuration of a general character reading device.

FIG. 28 is a diagram showing unevenness of a character boundary area after binarization processing due to a difference in resolution.

[Explanation of symbols]

I1 Scanner input multi-valued character image Separated image of I2 character boundary area I3 binary character image I4 Binarized image of only non-determined pixel area I5 Binary character image (before correction of character boundary area) P1 Undetermined pixel P2 fixed pixel T1 Threshold value used for separation processing of undetermined pixels T2 Threshold value in the neighborhood of undetermined pixels 1 Scanner section 2 Image storage 3 CPU (image processing unit) 4 OCR (character recognition processing unit)

─────────────────────────────────────────────────── ─── Continuation of front page (51) Int.Cl. ⁷ Identification code FI theme code (reference) H04N 1/403 H04N 1/40 103A 1/409 101C

Claims (16)

[Claims] 1. A scanner means for digitizing character information formed on a paper surface to take in multivalued image information, an arithmetic processing means for performing image processing arithmetic operation, the image information and an image processing result by the arithmetic processing means. An image processing device comprising: an image storage unit for storing the character boundary region; and a separation unit for separating a character boundary region generated by an input by the scanner unit; and a first binary for binarizing the character boundary region. And a second binarizing means for binarizing an area other than the character boundary area, and a correcting means for smoothing and correcting unevenness of the character boundary area after the binarizing processing. An image processing device characterized by: 2. The image processing apparatus according to claim 1, wherein the separating unit separates the character boundary area by using the information of the primary differential value and the density value as a feature amount. 3. The image processing apparatus according to claim 1, wherein the separating unit separates the character boundary area by an LBG algorithm using a first-order differential-density plane of an input image. 4. The binarizing means according to claim 1, wherein the first binarizing means binarizes the character boundary area using character density cross-section information generated by an input by the scanner means. Image processing device. 5. The image processing apparatus according to claim 1, wherein the second binarizing means performs binarization processing using a binarization processing result of the character boundary area in the vicinity. 6. The image processing apparatus according to claim 1, wherein the correction unit performs smoothing by detecting and filling irregularities generated in the character boundary area of the image after the binarization processing. . 7. The image processing apparatus according to claim 1, wherein the character boundary area is separated, and the binarization processing is switched for each of the separated areas. 8. The first binarizing means binarizes the character boundary area by using a character density section generated by an input by the scanner means and information on its shape. The image processing apparatus according to claim 1. 9. A scanner input multi-valued character image,
A separation step of separating a character boundary area generated by the scanner input, a first binarization step of binarizing the separated character boundary area, and a binarization processing of an area other than the character boundary area. An image processing method comprising: a second binarization step; and a correction step of smoothing and correcting the unevenness of the character boundary area after the binarization processing. 10. The image processing method according to claim 9, wherein in the separating step, the character boundary area is separated by using information of a first-order differential value and a density value as a feature amount. 11. The image processing method according to claim 9, wherein in the separating step, the character boundary area is separated by an LBG algorithm using a first derivative-density plane of an input image. 12. The image processing according to claim 9, wherein in the first binarization step, the character boundary area is binarized using character density cross-section information generated by the scanner input. Method. 13. The image processing method according to claim 9, wherein in the second binarizing step, the binarizing process is performed using a binarizing process result of the neighboring character boundary region. 14. The image processing method according to claim 9, wherein in the correction step, smoothing is performed by detecting and filling unevenness generated in the character boundary area of the image after the binarization processing. . 15. The image processing method according to claim 9, wherein the character boundary area is separated, and the binarization processing is switched for each of the separated areas. 16. The binarization step of the first binarization step is characterized in that the character boundary area is binarized by using a character density section generated by the scanner input and information on its shape. 9. The image processing method described in 9.