JPH08138050A - Color picture recognition method and object extraction method - Google Patents

Abstract

PURPOSE: To not only secure the information volume required for picture recognition but also compress the data volume to recognize an object with high precision at the time of recognition processing of a color picture. CONSTITUTION: A color picture signal conversion part l converts an inputted color picture signal into a chromaticity. A binarization processing part 2 compares this chromaticity with a preliminarily given threshold to generate a binary picture. A run length calculation part 3 obtains run length in the main scanning direction and subscanning direction from the generated binary picture, and a recognition part 4 performs the recognition processing of the color picture based on the distribution form of run length.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a color image recognition method and an object extraction method for extracting and recognizing a specific image from a color image and a monochrome image.

[0002]

2. Description of the Related Art It is expected that products for processing color images, such as color copying machines, color printers, color scanners and color image communication equipment, will increase more and more in the future. Color images have become easier to use than before due to advances in hardware, especially low cost and large memory capacity, and lower communication costs, but the amount of color image data is enormous (for example, , A3 size is 96 Mbytes), so that the same processing as a binary image cannot be performed at present.

In particular, image recognition (specific image recognition, OCR
In a technique requiring complicated processing such as (), the processing amount becomes enormous, and image recognition in a color image is more difficult to realize.

[0004]

Conventionally, as a method of identifying a specific color image, for example, each pattern portion forming an image has a distribution in a unique color space, and therefore appears in each pattern portion. Identify the distribution in the unique color space,
There is a method of extracting an image portion having the same characteristic as the specified characteristic (see Japanese Patent Laid-Open No. 4-180348). However, with this method, images that have the same spread in the color space cannot be identified even if the color distributions inside are different, that is, if the spreads in the color space are the same, the spread Images with different color distributions may be erroneously detected as specific images.

On the other hand, various methods have been proposed as a method of extracting an object required for recognition processing. For example, by extracting a black connected rectangle from an image and comparing it with a preset threshold value, the character There is an image extraction method for determining the rectangle of the above and the rectangle of the line figure (see Japanese Patent Laid-Open No. 162162/1988). This method does not identify the extracted line figure in more detail like a horizontal ruled line, a vertical ruled line, a table, a surrounding frame, etc., and cannot support the extraction of a rotated object.

An object of the present invention is to provide a color image recognition method for recognizing an object with high accuracy by compressing the data amount while ensuring the amount of information necessary for image recognition when performing color image recognition processing. An object is to provide a method for extracting an object.

[0007]

In order to achieve the above object, in the invention according to claim 1, the chromaticity is obtained from the input color image signal, and the chromaticity is compared with a predetermined threshold value.
Generating a value image, calculating white run and black run lengths from the generated binary image, and recognizing the color image based on the calculated white run length and black run length Is characterized by.

According to a second aspect of the present invention, the input color image signal is divided into a plurality of areas, and a color image is recognized for each of the divided areas.

According to the third aspect of the invention, the chromaticity is obtained from the input color image signal, the chromaticity is compared with a predetermined threshold value to generate a binary image, and the binary image is generated from the generated binary image. It is possible to extract the target object by calculating the lengths of the white run and the black run, creating a histogram of the white run length and the black run length, and collating with a dictionary of the extraction target object created in advance. It has a feature.

The invention according to claim 4 is characterized in that the object is extracted based on the position of the peak of the histogram.

According to a fifth aspect of the present invention, the object is extracted based on the peak position and the peak height of the histogram.

According to a sixth aspect of the present invention, the input color image signal is divided into a plurality of areas, and an object is extracted for each of the divided areas.

[0013]

In one embodiment, the input color image signal is converted into chromaticity for each pixel by the color image signal conversion section.
The binarization processing unit compares a threshold value given in advance with chromaticity to generate a binary image, and the run length calculation unit uses the generated binary image in the main scanning direction and the sub-scanning direction. The length is calculated, and the recognition unit recognizes the color image based on the run length distribution shape. In another embodiment, a run length histogram is determined and collated with a dictionary of objects to determine if the object is present in the color image.

[0014]

DESCRIPTION OF THE PREFERRED EMBODIMENTS One embodiment of the present invention will be specifically described below with reference to the drawings. <Embodiment 1> FIG. 1 shows the configuration of an embodiment of the present invention. In the figure, 1 is an input color image signal (RGB)
A color image signal conversion unit 2 for calculating chromaticity from a binary image processing unit 2 for generating a binary image by comparing a threshold value given in advance with the chromaticity. , A run length calculation unit for obtaining the length of the black run, a recognition unit 4 for recognizing a color image from the run length, and a control unit 5 for controlling each part.

The input color image signal (RGB) is
In the color image signal conversion unit 1, chromaticity (for example,
R / (R + G + B)), and the binarization processing unit 2 compares a threshold value given in advance with the chromaticity to generate a binary image. The run length calculation unit 3 calculates the length of the white run and the length of the black run in the main scanning direction and the sub scanning direction from the generated binary image, and passes the calculated length to the recognition unit 4. The recognition unit 4 recognizes an input color image based on the calculated run length distribution shape.

In the above embodiment, the chromaticity is taken as an example for explanation, but the present invention is not limited to this.
In addition to chromaticity, it may be a value obtained from the input signal such as an input signal (RGB), brightness, and saturation. Further, the threshold value may not be a fixed threshold value but may be a threshold value (function of the input signal) that changes depending on the value of the input signal.

<Embodiment 2> FIG. 2 shows the construction of another embodiment of the present invention. In this embodiment, a histogram creation unit 6 and a dictionary 7 of objects are added to the configuration of FIG. The histogram creation unit 6 creates a histogram for each white run and black run from the run lengths calculated by the run length calculation unit 3 of the first embodiment. In addition, for the image to be extracted, a histogram is created in advance by the same method as the input image and stored in the dictionary 7 of the object.

FIG. 3 is a diagram for explaining the comparison between the evaluation image and the dictionary of the object. As shown in FIG. 3, the collation in the recognition unit 4 is, for example, the frequency H1 of the run length L (= 50) of the white run and the frequency H of the run length Ld (= 50) of the dictionary 7.
Find the difference from 2 and calculate this for all run lengths (0 to L),
Or a run length within a specific range (for example, 50 to 10)
0) is summed. This sum (that is, the difference from the dictionary)
If is smaller than a predetermined value, the recognition unit 4 determines that the target object to be extracted exists. The recognizing unit 4 similarly performs the above-described processing on the black run, and determines whether or not an object is present in the input color image signal, together with the result of the white run.

In the above-mentioned embodiment, the difference in the frequency is used as the collation method, but the collation may be performed by using another method such as a correlation coefficient.

<Embodiment 3> In Embodiment 3, not all color images are targeted, but an image in which the image to be extracted is a repetitive pattern among color images (for example, the same pattern is repeated). Image). The configuration of the third embodiment is similar to that of the second embodiment.

In the case of such a repetitive pattern, a peak appears at a specific position in the frequencies of the run length histograms obtained in Examples 1 and 2. FIG. 4A is a diagram showing an example of a histogram when there is a specific pattern,
(B) is a figure showing an example of a histogram when there is no specific pattern. Assuming that the target pattern to be extracted is drawn with black pixels, the peak position of the black run corresponds to the line width, and the peak of the white run corresponds to the repeating width of the repeating pattern. Therefore, when the peak position obtained from the input image matches the peak position stored in the dictionary 7, it is determined that the target object to be extracted exists in the input image.

<Example 4> In Example 3 described above, attention was paid only to the position of the peak. In the fourth embodiment (this configuration is the same as that of the second embodiment), when it is desired to extract an object having a predetermined number or more of repetitive patterns, not only the peak position of the histogram but also its height (frequency). When there is a peak and the frequency is equal to or higher than a predetermined value, it is determined that the target object exists in the input image.

<Embodiment 5> When the above-described Embodiments 1 to 4 are applied to the entire image, if the object to be extracted exists only in a part of the image, run length histograms of other parts are displayed. As a result, the recognition accuracy of the object is lowered. Therefore, in the fifth embodiment, the input color image is divided into a plurality of areas, and the recognition processing of the first embodiment and the object extraction processing of the second to fourth embodiments are performed for each area, thereby performing the recognition processing. The accuracy of the extraction process of the object is improved. Further, when the size of the target object to be extracted is known in advance, the target can be extracted with higher accuracy by dividing the size into region sizes.

The color image recognition method and the object extraction method of the above-described first to fifth embodiments can perform high-speed processing because they do not require complicated calculation processing, but recognize the structural arrangement of repetitive patterns. Difficult to do. Therefore, when it is necessary to recognize the structural arrangement or the like, it is possible to recognize the structural arrangement or the like by performing more detailed recognition processing after the processing of the present invention.

[0025]

As described above, according to the color image recognition method and the object extraction method of the present invention, the input image data amount is converted into a run length histogram while retaining the information amount necessary for image recognition. By compressing, the object can be recognized at high speed. Further, it is possible to highly accurately extract the image of the repeating pattern from the input color image.

[Brief description of drawings]

FIG. 1 shows the configuration of an embodiment of the present invention.

FIG. 2 shows a configuration of another embodiment of the present invention.

FIG. 3 is a diagram illustrating matching between an evaluation image and a dictionary of an object.

FIG. 4A is a diagram showing an example of a histogram when there is a specific pattern, and FIG. 4B is a diagram showing an example of a histogram when there is no specific pattern.

[Explanation of symbols]

 1 Color image signal conversion unit 2 Binarization processing unit 3 Run length calculation unit 4 Recognition unit 5 Control unit

Claims (6)

[Claims] 1. A chromaticity is obtained from an input color image signal, a chromaticity is compared with a predetermined threshold value to generate a binary image, and a white run and a black run are generated from the generated binary image. A color image recognition method, wherein a length is calculated, and the color image is recognized based on the calculated white run length and black run length. 2. The color image recognition method according to claim 1, wherein the input color image signal is divided into a plurality of regions, and a color image is recognized for each of the divided regions. 3. A chromaticity is obtained from an input color image signal, the chromaticity is compared with a predetermined threshold value to generate a binary image, and a white run and a black run are generated from the generated binary image. Object extraction characterized in that the object is extracted by calculating the length, creating a histogram of the white run length and the black run length, and collating with a dictionary of the extraction object created in advance. Method. 4. The object extracting method according to claim 3, wherein the object is extracted based on the position of the peak of the histogram. 5. The object extracting method according to claim 3, wherein the object is extracted based on a peak position and a peak height of the histogram. 6. The object extracting method according to claim 3, wherein the input color image signal is divided into a plurality of areas, and the object is extracted for each of the divided areas.