CN101295359A - Image processing program and image processing device - Google Patents

Abstract

The invention provides an image processing program and an image processing device, recorded words, prints, marks and the like are extracted in a high accuracy from document images comprising color bias or fuzzy color. The image processing method of the invention is composed of: a background removing and data generating part for removing background parts from the color images or bright images, generating, and generating the background removing data which represents the parts other than the background; a profile color converting and data generating part for, in the parts other than the background in the color images or bright images, generating the data which converts the profile color of the parts other than the background into pixel color at inner side of the profile of the parts other than the background.

Description

Image processing program and image processing device

technical field

The present invention relates to a method of taking pictures of documents such as bills by using an optical character reading device (OCR: Optical Character Reader), a scanner, a digital camera, etc., and extracting images of specific objects such as words, imprints, marks, etc. from the generated document images. Processing method and image processing device.

Background technique

In financial institutions and local governments, scanner devices such as OCR are used to improve the efficiency of document processing tasks such as forms. The main functions of OCR are the generation of document images, the extraction of text in document images, and text recognition. Types of document images to be generated include binary images, luminance images, and color images.

Processing using binary images reduces processing time due to the small amount of data. However, in binary image processing, when preprinted ruled lines, bit lines, reminder characters, shadows and handwritten or later printed entry characters are largely overlapped. , it is difficult to distinguish them. Therefore, noise may be generated in the character extraction result, or part of the extracted character may be missing, resulting in a character recognition error.

Processing using a luminance image is monochrome shading image processing. Since preprinted and inscribed characters are differentiated by their different luminance values in luminance image processing, it is easier to distinguish them than binary image processing when preprinted and inscribed characters overlap. However, when the luminance values of preprinted and inscribed characters are similar, their discrimination accuracy becomes low.

In processing using a color image, since preprinted and inscribed characters can be distinguished according to the color difference, it is easier to distinguish them than brightness image processing. In color image processing, written characters, imprints, marks, etc. are extracted by removing preprinted colors.

In this method, there is a method of removing the color specified before inputting the form as in [Patent Document 3], and extracting the ruled lines in the input form as in [Patent Document 1] or [Patent Document 2]. Such a specific shape part is a method of removing the same color as the color of the extracted part.

【Patent Document 1】JP-A-2003-196592

【Patent Document 2】JP-A-2005-258683

【Patent Document 3】JP-A-2006-134355

【Patent Document 4】JP-A-2004-336106

[Patent Document 5] JP-A-2005-18810

In the above-mentioned color image processing, due to color deviation in the image generated by OCR, scanner, and digital camera, specific objects such as written characters or imprints cannot be correctly extracted, leaving a part of preprint, or a part of the specific object lack of issues.

The so-called color deviation refers to the positional deviation of at least one of the sensed color components of the three primary colors, the R value that becomes the red component, the G value that becomes the green component, and the B value that becomes the blue component. Causes of color shift include chromatic aberration of the lens, arrangement position of the sensor, conveyance speed, and the like. In particular, in a scanner using a two-dimensional CCD such as a desktop scanner or a digital camera, many color shifts due to chromatic aberration occur.

Due to the color shift, a false color different from the original color of the specific object occurs in the outline of the specific object such as preprinted or written characters. For example, red and blue false colors may be generated on the outline of black characters, or a reddish false color may be generated on the outline of blue ruled lines. Therefore, an error occurs in color image processing that distinguishes written characters from preprinted characters based on color information.

In this regard, [Patent Document 4] attempts to remove chromatic aberration of the lens, and [Patent Document 5] attempts to remove chromatic aberration caused by the arrangement position of the sensor.

However, it is difficult to completely remove color shift from an image even with a method of calculating and correcting the shift amount. In addition, a problem arises that it takes a lot of time for the grid line to correct the color misalignment with higher precision.

In addition, in the above-mentioned color image processing or luminance image processing, when color blurring occurs in the image, there is a problem that a specific object such as written characters or prints cannot be extracted correctly, a part of the preprint remains, or a part of the character is missing. .

The so-called blurred color means that the color of the ruled line or the outline of the written text is blurred, resulting in a light color. Due to the blurred color, the dispersion of color information such as red component, blue component, green component, lightness, chroma, hue, brightness, etc. of preprinted or inscribed text becomes larger, so it becomes difficult to distinguish between inscribed text and preprinted.

Contents of the invention

The present invention was made in view of these problems, and provides an image processing method and an image processing device for extracting specific objects such as written characters, prints, and marks with high precision from document images containing color deviation or color blur.

In order to achieve the above object, the present invention has the following features in the image processing method for extracting specific objects such as written characters, imprints, marks, etc. from color images or brightness images of documents such as ledgers read by a scanner or a digital camera, It has background removal generation processing for removing the background from a color image or a luminance image, and generating background removal data showing the portion other than the background; and generating a contour of the portion other than the background in the color image or luminance image. outline color conversion data generation processing of converting color information into data of color information of an image inside the outline of a portion other than the background; and specific object extraction processing of extracting a specific object portion.

In addition, the above-mentioned outline color conversion data generation process is characterized in that, for a pixel of interest in a color image or a luminance original image, a red component, a blue component, Color information such as green component, lightness, chroma, hue, and brightness is converted into low-brightness color expansion brightness data of the color information of the pixel with the lowest brightness value among nearby pixels and pixels of interest.

The specific object discrimination process described above is characterized in that ruled line extraction, specific object candidate extraction, ruled line color information and specific object color information estimation, and specific object discrimination are performed.

According to the present invention, even if it is a color image or a luminance image with color deviation or color blur, specific objects such as preprints, engraved characters, imprints, marks, etc. can be distinguished with high precision, for example, only engraved characters can be extracted with high accuracy. Not limited to writing text, it is also possible to extract specific objects such as imprints and marks in document images with high precision.

Description of drawings

FIG. 1 is a diagram showing the configuration of specific object extraction processing.

FIG. 2 is a diagram showing an image processing device.

Figure 3 is an example of a color image.

Figure 4 is the background removal data.

Fig. 5 is the discrimination result of a specific object.

FIG. 6 is a diagram showing an example of background removal data generation processing.

FIG. 7 is a diagram showing conventional specific object discrimination processing.

Figure 8 is the result of grid line extraction.

Figure 9 is the result of grid line removal.

Fig. 10 shows the results of specific object candidate extraction.

Fig. 11 is an example of color deviation of written characters.

Fig. 12 is an example of color deviation of ruled lines.

FIG. 13 is a diagram showing an example of outline color conversion data generation processing.

FIG. 14 is a diagram showing contour color conversion data generation processing of the image in FIG. 11 .

FIG. 15 is a diagram showing contour color conversion data generation processing of the image in FIG. 12 .

FIG. 16 is a diagram showing specific object discrimination processing.

FIG. 17 is a diagram showing a specific object discrimination process using only the estimation of the color of the ruled line.

FIG. 18 is a diagram showing a specific object discrimination process using only the estimation of the color of the specific object.

FIG. 19 is a diagram showing specific object discrimination processing using clustering.

FIG. 20 is a diagram showing the configuration of a specific object extraction processing program with color misalignment correction added.

FIG. 21 is a diagram showing the configuration of a specific object extraction processing program having an extraction object color designation function.

Fig. 22 is a diagram showing a subordinate specific object discrimination processing program including specifying an extraction target color.

FIG. 23 is a diagram showing specific object discrimination processing using a cluster including a specified extraction object color.

FIG. 24 is a diagram showing an example of color blur.

FIG. 25 is a diagram showing the state of contour color conversion data generation processing for an image with color blur.

FIG. 26 is a diagram showing an example of contour color conversion data generation processing for luminance image input.

Fig. 27 is a display example of a color image.

Fig. 28 is a display example of a determination result of a specific object.

(Symbol Description)

101 image acquisition module

102 background removal data generation module

103 contour color conversion data generation module

104 specific object discrimination module

105 control module

Detailed ways

Hereinafter, an image processing method and an image processing device to which the present invention is applied will be described in detail with reference to the drawings.

Example 1

FIG. 2 is a diagram showing an embodiment of an image processing device of the present invention.

This is an image processing device in which a communication device 201, an image acquisition device 202, a display device 203, an external storage device 204, a memory 205, a CPU (Central Processing Unit) 206, and an input device 207 such as a keyboard or a mouse are connected by a communication line such as a PCI bus. 208.

The program having the configuration of the specific object extraction process shown in FIG. 1 is stored in a storage device such as the external storage device 204 or the memory 205 , and is executed by the CPU 205 .

The color image or brightness image of the document input to the CPU may be input from the image acquisition device 202 such as a scanner or OCR or the communication device 201 , or may be stored in the external storage device 204 .

The results of the specific object extraction processing may be output to the display device 203 , output to the outside via the communication device 201 , or used in other programs in the image acquisition processing device 208 . As an example of other programs, there is a program for character recognition.

FIG. 27 shows an example of displaying a color image input from the image acquisition device 202 or the communication device 201 or stored in the external storage device 204 in a display window 2701 on the display device 203 . In addition, FIG. 28 is an example in which the result of the specific object extraction process is displayed in the display window 2702 on the display device 203 .

FIG. 1 is a diagram showing the structure of a specific object extraction processing program to which the present invention is applied. The specific object extraction processing program is composed of an image acquisition module 101 , a background removal data generation module 102 , an outline color conversion data generation module 103 , a specific object discrimination module 104 , and a control module 105 .

The image acquisition module 101 performs image acquisition processing for acquiring a color image or a brightness image obtained by imaging a paper document or the like by using a scanner or OCR.

The background removal data generation module 102 performs background removal data generation processing for generating background removal data from a color image or a luminance image input to the CPU 206 .

For example, when obtaining a color image of preprinted and written characters 304 including ruled lines 301, bit lines 302, and shadows 303 as shown in FIG. bit lines and write data into the text section.

The background removal data generation process is a process of removing the background part in the image, and extracting the preprinted part and the written character part of the ruled line and the bit line. There are various methods for realizing this processing, and the method shown in FIG. 6 is adopted.

First, in the luminance value data generation process 601, a luminance image represented by luminance is generated from a color image represented by the three primary colors of RGB (R value, G value, and B value). Then, in a block generation process 602, the luminance image is divided into a plurality of blocks. Finally, in the binarization process 603 , binary data is generated for each block in which pixels with low brightness values are set to black and pixels with high brightness values are set to white. The binary data thus generated is, as shown in FIG. 4 , background removal data in which black pixels represent parts other than the background.

The outline color conversion data generation module 103 performs outline color conversion data generation processing for generating outline color conversion data 1303, which is an input color image 604 and background removal data 605, and ruled lines, bit lines, and written characters. The color of the contour is converted to the color of the inner part of the contour. Additionally, color image 604 may be a luminance image.

The specific object discrimination module 104 performs a specific object judgment process for the background removal data 605 input to the CPU 206, refers to the outline color conversion data 1303, and generates data representing the written text portion shown in FIG. 5, and outputs the specific object judgment result 706.

Here, conventional specific object discrimination processing will be described with reference to FIG. 7 . In conventional specific object discrimination processing, background removal data is input, a color image is referred to, and a specific object discrimination result is output.

FIG. 7 shows conventional specific object discrimination processing. First, in the ruled line extraction process 701', the ruled line portion is extracted. In this process, a ruled line portion is extracted by extracting a portion where black pixels in the background removal data are connected linearly for a long period of time. The result is Figure 8.

Then, in the ruled line removal process 702', the ruled line removed data from which the ruled line portion has been removed from the background removed data is generated. The result is Figure 9.

Then, in the specified object candidate extraction process 703', from the ruled line removal data, the written character portion candidates that are candidates for the written character portion to be specified are extracted using the size and position information of the rectangle. The result is Figure 10.

Then, in the ruled line color and specific object color estimation process 704', by referring to the color image 604, the color information of the ruled line part which is the color information of the ruled line part and the entry of the color information which is the candidate character part are estimated. The color information of the alternate part of the text.

Then, in the specific object discrimination process 705', it is judged whether or not each pixel in the black pixel portion in the background removal data is a pixel of written text by using the ruled line portion color information and the written text candidate portion color information. This processing is a process of discriminating whether the color information of the color image belongs to the color information of the ruled line part or the color information of the write-in character candidate part at each pixel position of the black pixel part in the background removal data.

Specifically, the following processing is performed for each pixel in the black pixel portion in the background removal data 605 . In the processing of a certain black pixel position (Xa, Xb) in the background removal data 605, it is judged that the color information in (Xa, Xb) of the color image 604 is different from the estimation processing 704' of the ruled line color and the specific object color. Which one of the color information of the output ruled line part and the color information of the candidate character part to be output is close to. And, if the color information of (Xa, Xb) is close to the color information of the ruled line part, then it is determined that the position of (Xa, Xb) is the ruled line part, if the color information of (Xa, Xb) is close to the color information of the candidate part of the text, Then it is judged that the position of (Xa, Xb) is the part of writing characters.

As the color information of the color image 604, the R value, G value, and B value of the three primary colors of RGB can be used, or color information converted from them, such as brightness value or hue, chroma, and lightness in HSV space. In addition, only one of these values may be used, or a plurality of values may be used. Furthermore, in the discriminant method, various discriminant algorithms using teacher data can be used. For example, a neural network, a linear recognizer, Mahalanobis Distance, etc. are used.

Then, by referring to the color image 604, the specific object determination process is performed, the specific object determination result 706' is output, and the specific object determination process ends.

However, in the case of conventional specific object discrimination processing, if there is color deviation in the input color image 604, the color information in the color image 604 also has deviation, so based on the color information of the part close to the ruled line and the color of the candidate part of the written character There is a possibility that the judgment itself of which color information among the information is biased, and therefore there is a problem that the color information cannot be used to distinguish between preprinted characters and inscribed characters. Therefore, the judgment result of the present invention, such as the output shown in FIG. 5 , may not be obtained from the specific object judgment result obtained by the conventional method.

Here, FIG. 11 is an example of an image (written characters) with color deviation. In FIG. 11 , a blue false color and a red false color are generated on the outline of a written character that is originally black.

In addition, FIG. 12 is also an example of an image (ruled lines) with color shift. In Figure 12, a reddish false color is produced on the outline of the originally blue grid. Consider a case where only the typed characters are extracted using color information from an image including black typed characters and blue ruled lines.

When there is no color shift between the written characters and the ruled lines, only the written characters can be extracted by conventional specific object discrimination processing. However, in images with color shift as shown in Figures 11 and 12, since there is a red component in both the outline of the written characters and the outline of the ruled lines, the outline of the ruled lines may be noise or part of the text lack of situation. When color shift occurs in this way, there is a problem that preprinted and written characters cannot be distinguished using color information.

11 and 12, the image processing apparatus to which the present invention is applied performs contour color conversion data generation processing, and performs specific object discrimination processing with reference to the contour color conversion data.

The above-mentioned outline color conversion data generation module 103 performs outline color conversion data generation processing. Specifically, in the portion other than the background in the color image 604 , data in which the color of the outline of the portion other than the background is converted into the color of the pixel in the inner side of the outline of the portion other than the background is generated. That is, data is generated by converting the color information of the outline of the ruled lines, bit lines, and written characters in the color image into the color information of the inner portion of the outline.

FIG. 13 is an example of a specific processing flow of outline color conversion data generation processing.

In the contour color conversion data generation process, a color image 604 and background removal data 605 are input from the communication device 201 , the image acquisition device 202 , or the external storage device 204 via the memory 205 .

In addition, in the nearby luminance value generation process 1301 and the low luminance color expansion process 1302, pixels in areas other than the background in the color image are selected one by one (the selected pixels are referred to as pixels of interest), and the color information of the pixel of interest is converted. . These two processes are repeated until all pixels in areas other than the background in the color image are processed.

In the neighborhood luminance value generating process 1302, pixels in a region surrounding the pixel of interest (in the example of FIG. 14 , 9 pixels in a region of a 3×3 range centered on the pixel of interest) are respectively generated. Brightness value. The pixels other than the pixel of interest in the area are assumed to be nearby pixels. The above-mentioned range is not limited to 3×3, and may be 2×2 or 4×4, for example. In addition, the pixel of interest is not limited to the center within the domain, and the domain may be set such that the pixel of interest is located at any position within the domain.

Then, in the low-brightness color expansion process 1302, the color information (for example, R value, G value, and B value) of the pixel of interest is converted into color information of a pixel having the lowest luminance value among the pixel of interest and nearby pixels. In this way, converting the color information of the contour portion where the R value, G value, and B value deviates into color information inside the contour portion is a process of converting the false color into the original color information in an analog manner.

More specifically, the luminance values of the pixel of interest and nearby pixels in the area are calculated, the pixel with the lowest luminance value is extracted, and the color information of the pixel of interest is converted into the color information of the pixel with the lowest luminance value. If the luminance value of the pixel of interest is the lowest luminance value, the color information of the pixel of interest is maintained as it is. In this way, outline color conversion data 1303 is generated as data in which color information has been converted for ruled lines, bit lines, and written characters in the color image 604 .

In the outline color conversion data generation process, for example, in the case of writing characters, as shown in FIG. It is black with low luminance inside the outline.

Also, with the outline color conversion data generation processing, for example, in the case of a ruled line, as shown in FIG. Blue with low brightness inside this outline.

FIG. 16 is a diagram showing a specific processing flow of specific object discrimination processing in the first embodiment.

First, a ruled line extraction process 701 is performed in which background removal data is input and a ruled line portion is extracted.

Next, ruled line removal processing 702 is performed to generate ruled line removed data in which ruled line portions are removed from the background removed data.

Then, specific object candidate extraction processing 703 is performed for extracting written character portion candidates that are candidates for written character portions to be specified from the ruled line removal data using information on the size and position of a rectangle.

Then, in the specific object discrimination process of the present invention, the RGB values of the outline color conversion data 1303 are referred to in the ruled line color and specific object color estimation process 1601 and the specific object discrimination process 1602 .

In the area of the outline color conversion data 1303 corresponding to the black pixel area of the background color removal data, since there are fewer pixels having false colors due to color shift, the estimation accuracy of the specific object color and ruled line color is better, as As a result, the accuracy of discrimination between ruled lines and specific objects is also improved.

In this way, in the image processing device 208 to which the present invention is applied, in the estimation 1601 of the color of the ruled line and the color of the specific object and the discrimination process 1602 of the specific object, it is possible to set the written character part to black and the ruled line part to be blue. It is processed by color, so it is possible to correctly identify the part of the written character.

As described above, according to the image processing device 208 , the outline color conversion data including the RGB values after the outline color conversion processing is referred to, so it is possible to accurately extract the writing character to be specified from the color image including color deviation. In addition, the character recognition device that receives the input character extraction result as the output of the image processing device as input can obtain a higher-precision recognition result. In addition, the extraction of written characters is used as an example, but it can also be extracted with high accuracy in the case of extracting imprints or marks.

Next, other embodiments of the present invention will be described.

Example 2

As shown in FIG. 17 , in the specific object determination unit 104 , specific object extraction processing may be employed in which specific objects are identified only by estimating the ruled line color.

In the processing shown in FIG. 17 , only the color information of the ruled line is estimated by referring to the outline color conversion data in the ruled line color estimation process 1701 . Then, in the ruled line color part removal process 1702, the ruled line color part is removed from the background removal data 605 by using the ruled line color information, and the writing character part to be specified is determined.

Example 3

As shown in FIG. 18 , in the specific object discrimination unit 104 , specific object extraction processing may be employed in which specific object discrimination processing is performed using only the estimation of the color of the specific object.

In the process shown in FIG. 18 , only the color information of the specific target candidate is estimated by referring to the outline color conversion data 1303 in the specific target color estimation process 1801 . Then, in the extraction process 1702 of the color portion of the specific object, the written character portion to be the specific object is extracted from the background removal data 605 using the color information of the specific object.

Example 4

As shown in FIG. 19 , specific object extraction processing may be employed in which specific object identification processing is performed using clustering in the specific object identification unit 104 .

In the processing shown in FIG. 19 , the result of ruled line extraction is not used, but only the color information of parts other than the background is used for discrimination. First, in clustering processing 1901, the outline color conversion data 1303 of parts other than the background are clustered. In clustering, the R value, G value, and B value of the RGB 3 primary colors can be used, or the color information converted from them, such as the brightness value or the hue, chroma, and lightness of the HSV space. In addition, only one of these values may be used, or a plurality of values may be used. Among the clustering methods, there are methods such as k-means method, region expansion method or discriminant analysis.

Then, in the selection process 1902 of the class of the specific object, the class of the specific object is selected from the plurality of classes obtained by clustering. There are various methods of selection, for example, a method of selecting a class having a high brightness value.

Then, in the extraction 1903 of the color portion of the specific object class, pixels having the color information of the selected class are extracted from the black pixel portion of the background removal data, and the writing characters to be specified are extracted.

Example 5

It is also possible to employ a specific object extraction process in which a color misalignment correction module 2001 is added to the configuration of the specific object extraction processing program shown in FIG. 1 .

This specific object extraction processing program is configured as shown in FIG. 20, and performs the same processing as the embodiment shown in FIG. 1 except for the processing described below.

The color misalignment correction module 2001 executes color misalignment correction processing. The color shift correction process generates color shift correction data that reduces color shift by changing the R value, G value, and B value of the color image 604 acquired by the document image acquisition process, or by expanding or reducing it.

And, in contrast to the configuration shown in FIG. 1, the data input to the background removal data generation process and the outline color conversion data generation process use the color image 604. In the embodiment in FIG. 20, the data input to the background removal data generation process The data in the processing and the contour color conversion data generation processing are color misalignment correction data. In this way, specific objects such as written characters can be extracted with high precision even in an image with a large amount of color shift.

Example 6

It is also possible to employ the specific object extraction processing in which the specified color acquisition module 2101 is added to the configuration of the specific object extraction processing program shown in FIG. 1 .

This specific object extraction process has a configuration as shown in FIG. 21, and performs the same processing as in the embodiment shown in FIG. 1 except for the processing described below.

In the designated color acquisition module, designated color acquisition processing is performed. In the specified color acquisition process, the specified extraction target color information 2203 that is the color specified as the specific object to be extracted is acquired. The specified extraction target color information includes information previously specified by the user in a program, information input from an input device such as a keyboard or a mouse, and the like. In addition, the color information may use the R value, G value, and B value of RGB, or may be color information converted from them, such as brightness value or hue, chroma, and lightness in HSV space. In addition, only one of these values may be used, or a plurality of values may be used. In addition, it may be a value for displaying one color, or a range of values for displaying a color.

Furthermore, the specific target discrimination process is a process in which the specified extraction target color information 2203 is included in the input as shown in FIG. 22 or FIG. 23 .

In FIG. 22 , in the specific object discrimination process in FIG. 16 , specific object discrimination 2201 is performed using the specified extraction target color information 2203 and the result of estimation 1601 of the ruled line color and the specific object color.

In FIG. 23 , in the specific object discrimination process using the clustering 1901 in FIG. 19 , the class of the specific object is specified 2301 using the specified extraction object color information 2203 .

The above-described embodiments are effective not only for the problem of color shift but also for the problem of color blur. Fig. 24 is an example in which the outline of the written characters is blurred with a light color. For the image in FIG. 24, when the outline color conversion data generation process is performed, the outline color conversion data 1303 shown in FIG. 25 is generated. In the outline color conversion data 1303, blurred light-colored parts in the input color image 604 are converted into dark colors. In this way, a specific object can be extracted with high precision even in an image with blurred colors.

In addition, the above-described embodiments are effective not only when a color document is input, but also when a luminance image that causes color blur is input. In the case where a luminance image is input, it can be handled by setting the outline color conversion data generation process to the process shown in FIG. 26 in the embodiment of FIG. 1 .

In FIG. 26 , a luminance image 2604 and background removal data 605 are input, and in the portion of the luminance image 2604 other than the background, the low-luminance color expansion process 2601 of the luminance image is performed on a pixel-by-pixel basis. In addition, the outline color conversion data 1303, which is data obtained by converting the luminance value in the luminance image 2604, is generated for the portion other than the background, that is, the ruled line, the bit line, and the written text.

In low-brightness color expansion processing 2601 of a brightness image, the brightness value of the pixel with the lowest brightness value among the pixel of interest and nearby pixels is converted into the brightness value of the pixel of interest.

Claims (15)

1. An image processing device, characterized in that, possesses: Remove the background from the input image information to generate background removal data representing regions other than the background; means for generating color conversion data that converts color information of pixels of an outline related to a specific object in an area other than the background of the input image information into colors of color information of pixels inside the aforementioned outline convert data; a unit storing the above-mentioned background removal data and the above-mentioned color conversion data; and A specific object discrimination unit that selects a specific object candidate from the background removal data, refers to the color conversion data, and outputs a specific object portion. 2. The image processing device according to claim 1, characterized in that: The image information described above is color image information or brightness information of an image. 3. The image processing device according to claim 1, characterized in that: In the above unit that generates color conversion data, selecting an area including a plurality of pixels from the background removal data representing an area other than the above-mentioned background, Generate brightness values for pixels within the above field, Color conversion data for converting color information of a pixel of interest within the above-mentioned area into color information of a pixel having the lowest luminance value among pixels within the above-mentioned area is generated. 4. The image processing device according to claim 1, characterized in that: The above-mentioned image information is any one of information stored in an external storage device, information acquired by an image acquisition device, or information input from a communication device. 5. The image processing device according to claim 1, characterized in that: The above-mentioned unit for generating color conversion data, Setting the pixel of interest on the above-mentioned image information, referring to nearby pixels that are pixels in a region surrounding the pixel of interest, Low-brightness color dilated luminance data in which the above-mentioned color information of the pixel of interest is converted into color information of a pixel having the lowest luminance value among nearby pixels and the pixel of interest is generated. 6. The image processing device according to claim 1, characterized in that: The above color information is at least any one of red component, blue component, green component, lightness, chroma, hue, and brightness. 7. The image processing device according to claim 1, further comprising: A unit for extracting grid line information from the above-mentioned background removal data; removing elements of grid line information extracted from the above-mentioned background removal data; and A specific object discrimination unit that selects a specific object candidate from the background removal data from which the ruled line information has been removed, refers to the color conversion data, and outputs the specific object portion. 8. An image processing method, utilizing a processing device with a storage unit, an image acquisition unit, a computing unit, and a display unit, the image processing method has: A step of removing the background from the image information input by the image acquisition unit, and generating background removal data representing an area other than the background; A step of generating color conversion data of converting color information of pixels of an outline related to a specific object in an area other than the background of the input image information into colors of color information of pixels inside the aforementioned outline convert data; a step of storing the above-mentioned background removal data and color conversion data in a storage unit; and A specific object candidate is selected from the background removal data stored in the storage means, the color conversion data is referred to, and a portion of the specific object is output to the specific object discrimination step in the display means. 9. The image processing method according to claim 8, characterized in that: The image information described above is color image information or brightness information of an image. 10. The image processing method according to claim 8, characterized in that: In the above step of generating color conversion data, selecting an area including a plurality of pixels from the background removal data representing image information other than the above-mentioned background, Generate brightness values for pixels within the field, Color conversion data for converting the luminance value of the pixel of interest within the area into the lowest luminance value among the pixels within the area is generated. 11. The image processing method according to claim 8, characterized in that: The above image information is any of information stored in an external storage device, information acquired by an image acquisition device, or information input from a communication device. 12. The image processing method according to claim 8, characterized in that: In the above step of generating color conversion data, For a pixel of interest within the domain, referring to nearby pixels that are pixels other than the pixel of interest within the domain, Low-brightness color dilated luminance data in which the above-mentioned color information of the pixel of interest is converted into color information of a pixel having the lowest luminance value among nearby pixels and the pixel of interest is generated. 13. The image processing method according to claim 8, characterized in that: The above color information is at least any one of red component, blue component, green component, lightness, chroma, hue, and brightness. 14. The image processing method according to claim 8, further comprising: A step of extracting grid line information from the background removal data; a step of removing gridline information extracted from the above-mentioned background removal data; and Select a specific object candidate from the background removal data from which the above-mentioned ruled line information is removed, refer to the above-mentioned color conversion data, and output the specific object discrimination step of the specific object portion. 15. An image processing program that causes a computer to function as a unit for image processing: A unit that removes the background from the image information input by the image acquisition unit, and generates background removal data representing regions other than the background; A unit for generating color conversion data that converts color information of pixels of an outline related to a specific object in an area other than the background of the input image information into color information of pixels on the inner side of the aforementioned outline convert data; means for storing the above-mentioned background removal data and color conversion data in the storage unit; and A specific object candidate is selected from the background removal data stored in the storage means, the color conversion data is referred to, and a portion of the specific object is output to the specific object discrimination means in the display means.

Images