JP2954218B2 - Image processing method and apparatus - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method and apparatus for extracting a feature amount of a contour line of character image data.

[Prior Art] Conventionally, as a method of extracting features of binary character image data,
The inclination of the outline is coded in four directions at the outline points of the character image data, and the direction codes are obtained for all the outline points. Then, for each block obtained by dividing the image area into, for example, 4 × 4 blocks, a histogram of the direction codes of the contour points included in the block is calculated to obtain 4 × 4 × 4 blocks.
There is a feature extraction method in which (= 64) direction code histograms are calculated, and these are used as feature amounts of character image data.

[Problems to be Solved by the Invention] In this case, in order to obtain a histogram of the direction code for each block, the image area is raster-scanned sequentially from the upper left of the area corresponding to the block, and hits the contour points of the character image. Each time, the direction code of the contour point is counted as that of the block. Further, as a device for preventing unnecessary scanning of a portion having no contour point, a method of storing the leftmost edge and the rightmost edge of the contour point for each scanning line and performing scanning only on that pixel is also considered. Was. However, in the above-described conventional example, an area including most non-outline pixels is scanned in order to obtain a direction code histogram for each block, which is extremely inefficient.

The present invention has been made in view of the above conventional example, and creates contour point data by combining coordinate information of a contour point and characteristic information of the contour point, and generates coordinate data of the contour point data and character image data. Provided is an image processing method and apparatus for comparing a coordinate value of each block to obtain a block in which the contour point is present, obtaining a characteristic amount of the block, and extracting a characteristic amount of the entire character image. The purpose is to do.

[Means for Solving the Problems] In order to achieve the above object, an image processing method of the present invention includes the following steps. That is, the contour of the character image data is tracked, and the features of the contour points that constitute the contour are extracted. When the features of the contour points are extracted, the coordinate information of the extracted contour points, The contour point data is created by combining the feature information of the contour points, the coordinate values of each contour point of the created contour point data are sequentially read, and the coordinate values of the character image data are divided into a plurality of blocks and compared with the coordinate values of each block. Then, a block in which a contour point corresponding to the contour point data exists is obtained, and a process of integrating the characteristic amount count value of the block in which the contour point exists exists is performed on all the contour points of the character image data. And the feature amount count value of the block is used as the feature amount of the character image data.

In order to achieve the above object, the image processing apparatus of the present invention has the following configuration. That is, when a contour point feature extraction unit that tracks a contour line of character image data and extracts a feature of a contour point constituting the contour line, and a contour point feature is extracted by the contour point feature extraction unit, Contour point data creating means for creating contour point data by combining coordinate information of the extracted contour points and feature information of the contour points; coordinate values of the contour point data created by the contour point data creating means Reading means for sequentially reading the coordinate values, and comparing the coordinate values read by the reading means with the coordinate values of each block obtained by dividing the character image data into a plurality of blocks, wherein there is a contour point corresponding to the contour point data. Block detecting means for obtaining a block which is present, and a character image which is a feature quantity of the character image data by integrating a feature quantity count value of the block obtained by the block detecting means. And having a data feature extraction means.

[Operation] With the configuration described above, the contour of the character image data is tracked, and the features of the contour points constituting the contour are extracted. When the features of the contour points are extracted, the extracted contour is extracted. Creates the contour point data by combining the coordinate information of the point and the characteristic information of the contour point, sequentially reads out the coordinate values of each contour point of the created contour point data, and divides the character image data into a plurality of blocks. The block in which the contour points corresponding to the contour point data are found by comparing the coordinate values of the contour point data, and the process of integrating the feature quantity count value of the block in which the contour points are present is performed on all of the character image data. , And operates so that the feature amount count value of those blocks is used as the feature amount of the character image data.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.

[Explanation of Functional Configuration (FIG. 1)] FIG. 1 is a block diagram showing the functional configuration of the character recognition device of the embodiment.

In the figure, reference numeral 100 denotes, for example, a document manuscript read photoelectrically.
This is an image memory for inputting and storing character image data of a value digital signal. However, the image memory 100 may be a storage device such as a disk that already stores character image data. 101 is a contour tracking unit that tracks the contour points of the detected character, and 102 is a contour in which a set of the contour point code of each contour point obtained by tracking the contour of the character image and the coordinates of the contour point are sequentially arranged. This is a contour point code array storage unit that creates and stores a point code array.

103 is based on the coordinate value of each contour point and the direction of the next contour point,
This is a contour point code calculation unit that calculates a contour point code. Ten
Reference numeral 4 denotes a block dividing unit for dividing the character image data into blocks, and reference numeral 105 denotes a belonging block determining unit for sequentially reading the contour point code array and determining the block to which the contour point of each array element belongs. Reference numeral 106 denotes a block-based contour point classification unit that counts the number of array elements for each block and contour point code, and creates a block-based contour point code histogram.

 Hereinafter, the operations of these units will be described in detail.

The character image data of the image memory 100 is composed of one byte per pixel, and the character part (black) is the least significant bit (LS).
B) is "1", and the least significant bit of the white part is "0". Also, the least significant bit (MSB) is a bit that is set to “1” when the contour has been tracked and has already been tracked.
By checking the value of this MSB, the same contour point
Can be prevented from being tracked.

The contour tracking unit 101 raster scans the image data in the image memory 100 from the lowest address and checks the character image data. The previous pixel is white, the current pixel is black, and the contour tracking is not performed (MSB is “0”). The pixel of ") is searched, and when the pixel is found, the first point is set as the contour start point.

 FIG. 3 is a diagram showing the principle of contour tracking.

Here, contour points are tracked in 3 × 3 pixel matrix units, and “*” marks indicate the current contour points, and “1” to
“7” indicates peripheral pixels. Now, assuming that the previous contour point is a pixel indicated by “1”, starting from the next pixel that is one step over, as a starting point, it is determined whether the pixels are white or black in the order of pixels “3” → “4”. The determination is made, and the first black pixel encountered is the next contour point.

Next, the operation of the contour point code calculation unit 103 will be described.
When the next contour point is obtained by the contour tracking unit 101 as described above, a contour point code is determined using a direction connecting the current contour point and the next contour point as a direction code. This direction code is coded corresponding to four directions, as shown in FIG. 4, and can take four values "0" to "3". The code obtained by the contour point code calculation unit 103 in this way is transferred to the contour point code array storage unit 102 via the contour tracking unit 101.
Is stored.

In this way, successively following the newly determined next contour point as the current contour point, the contour tracking is sequentially performed, and a code array having a set of a contour point code and a contour point coordinate as an element is a contour point code. It is created in the array storage unit 102.

The contour point code array storage unit 102 stores contour point codes and contour point coordinates of the contour points (the main scanning direction of the character image data is the X axis,
An array R [i] (hereinafter, referred to as a contour point code array. R [i] .code: contour point code, in which a set of an X coordinate value and a Y coordinate value when the sub-scanning direction is the Y axis is an element. R [i] .x: X coordinate, R [i] .y: Y coordinate) store the contour point code and the contour point coordinates of the current contour point.

The block dividing section 104 divides the inside of the image memory 100 and the inside of the character frame of character image data into, for example, 4 × 4 blocks as shown in FIG. Thus, when the character image data is divided into blocks, the belonging block determination unit 10
5 determines which block each contour point exists in based on the coordinate values of each contour point stored in the contour point code array storage unit 102. This is done by storing the coordinate value of the upper left point and the coordinate value of the lower right point of each block divided into blocks, and comparing these coordinate values with the coordinate values of each contour point.

The block-specific contour point classifying unit 106 sequentially reads the contour point code array R [i] of the contour points included in the block from the contour point array code storage unit 102, counts each block, and finally calculates the block-specific direction. Code histogram H [k]
[J] (k: block number 0-15, j: direction code 0-3)
Ask for.

[Explanation of Character Recognition Apparatus (FIG. 2)] FIG. 2 is a block diagram showing a schematic configuration of the character recognition apparatus of the embodiment which performs the above-described operation.

A CPU 201 outputs various control signals and controls the entire apparatus according to a control program and various data shown in the flowchart of FIG. Two
An image memory 02 stores character image data (image data) input from the scanner 205. 204 is CPU
This RAM is used as a work area 201 and temporarily stores various data. The RAM 204 includes a code array memory 204-1 for storing the contour point code array R [i] described above, a histogram memory 204-2 for creating a histogram for each block, and the like.

Reference numeral 208 denotes a scanner controller (SCTR) that controls the scanner 205 in accordance with instructions from the CPU 201. Reference numeral 205 denotes a scanner that reads a character original under the control of the SCTR 208 and inputs a digital signal. And a recognition result, and a display such as a CRT for displaying various messages to the operator. A CR 209 controls the display 206.
T controller (CRTC). Reference numeral 207 denotes a printer that outputs a recognition result and the like, and 210 denotes a printer controller (PRTC) that controls the printer 207.

[Explanation of Operation (FIGS. 1 to 6)] FIG. 6 is a flowchart for explaining a feature extracting operation in the character recognition device of the embodiment. Hereinafter, a control operation will be described based on this flowchart. Before the start of the operation, it is assumed that character image data is already input and stored in the image memory 202 from the scanner 205 or another device.

In step S1, the character image data is read from the image memory 202, a character frame including the character image is cut out, and raster scanning is sequentially performed from the lowest address of the image data, where the previous pixel is white, the current pixel is black and the outline is black. Pixels for which tracking has not been performed are searched, and in step S2, it is determined whether or not the starting point has been found. When the start point is found, the process proceeds to step S3, and a contour point code is obtained based on the positional relationship between the contour point and the previous contour point.

Then, the process proceeds to step S4, and the contour point code is stored in the code array memory 204-1 corresponding to the coordinate position of the contour point. Then, in step S5, the next contour point is detected, and in step S6, it is determined whether the contour point matches the contour point (contour start point) obtained in step S1. If not coincident with the start point, the flow returns to step S3 to execute the above-described operation. If it coincides with the start point, that is, if a closed contour is detected, the process returns to steps S1 and S2 to check whether another contour start point exists. When all the contour lines for the character image data in the character frame thus cut out are detected, the process proceeds to step S7.

In step S7, the character image data in the image memory 202 is divided into, for example, 4 × 4 blocks. In step S8, the outline point code array is read out from the code array memory 204-1. In step S9, those outline points are stored in any block. Determine if it exists. As described above, the coordinate value of the upper left corner and the coordinate value of the lower right corner of each block are compared with the coordinate value of each contour point to determine whether or not the contour point exists in the block.

Thus, in step S10, the contour point code R [i] .code
+1 is added to the block-specific directional code histogram H [m] [R [i] .code] of the histogram memory 204-2 corresponding to the block number m (m is 0 to 15) belonging to When the processing for all the outline point code arrays R [i] stored in the code array memory 204-1 is completed in step S11, the block-specific directional code histogram H [m] [j as the characteristic amount of the character image data is obtained. ] (J is the direction code) is completed.

As described above, according to this embodiment, the contour point code is obtained at the time of contour point tracking and stored in a code array format, so that the character image data can be scanned by one raster scan for contour tracking. Since only scanning is required and scanning for creating a block-specific direction code and a histogram is not required, there is an effect that the feature extraction of character image data can be performed at high speed.

[Other Embodiment (FIG. 7)] FIG. 7 is a block diagram for explaining another embodiment.
Parts common to FIG. 1 are indicated by the same symbols.

Here, an example in which a block-specific contour point histogram is obtained as a feature value is described. Here, the number of contour points included in each block may be obtained for each block.
The contour point code calculation unit 103 is not required as compared with the case of FIG. Further, the contour point code array storage unit 102 'only needs to store only the array having only the coordinates of the contour points as elements. Therefore, in this case, in step S10 in FIG. 6, that is, in the block-specific contour point classifying unit 106 ', it is only necessary to increment the block-specific contour point histogram H [m] by 1 for the block number m.

As described above, according to the present embodiment, the processing time is greatly reduced as compared with the case where the character image data is raster-scanned again to extract features, and the ratio of the processing time is approximately (number of contour pixels) / (Total number of pixels), which is about 1/10
It will be about 0 to 1/10.

[Effects of the Invention] As described above, according to the present invention, when extracting a feature amount of character image data, if a feature of an outline point of the character image data is extracted, a block of the block including the outline point is extracted. Since only the feature amount count value needs to be incremented, it is not necessary to perform raster scan on all the coordinates constituting the character image data, including the coordinates where there is no contour point, to check, and it is necessary to extract the feature amount of the character image data. There is an effect that processing can be performed efficiently and in a short time.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a functional configuration of the feature extraction of character image data of the embodiment, FIG. 2 is a block diagram showing a schematic configuration of a character recognition device of the embodiment, and FIG. FIG. 4, FIG. 4 is a diagram showing an example of a direction code at an outline point, FIG. 5 is a diagram showing an example of block division of character image data, and FIG. 6 is a flowchart showing a feature extraction process in the character recognition device of the embodiment. FIG. 7 is a functional block diagram showing a feature extracting method according to another embodiment. In the figure, 100: image memory, 101: contour tracking unit, 102:
... Contour point code array storage section, 103... Contour point code calculation section, 104... Block division section, 105... Belonging block determination section, 106.
…… Image memory, 203… ROM, 204… RAM, 204−
1. Code array memory, 204-2 Histogram memory, 205 Scanner, 206 Display, 207
... It is a printer.

Claims (2)

(57) [Claims] 1. A contour line of character image data is traced to extract features of contour points constituting the contour line. When the features of the contour point are extracted, coordinate information of the extracted contour point is extracted. And the feature information of the contour points are combined to create contour point data, the coordinate values of each contour point of the created contour point data are sequentially read, and the coordinates of each block obtained by dividing the character image data into a plurality of blocks The block in which the contour point corresponding to the contour point data is present is determined by comparing the value of the block with the contour point data, and the process of integrating the feature count of the block in which the contour point is present is performed on all the contours of the character image data An image processing method executed for a point, wherein a feature amount count value of the block is used as a feature amount of the character image data. 2. A contour point feature extracting means for tracing a contour line of character image data and extracting features of contour points constituting the contour line, wherein the contour point feature is extracted by the contour point feature extracting means. In this case, contour point data creating means for creating contour point data by combining coordinate information of the extracted contour points and feature information of the contour points, and contour point data created by the contour point data creating means Reading means for sequentially reading the coordinate values of the above, and comparing the coordinate values read by the reading means with the coordinate values of each of the blocks obtained by dividing the character image data into a plurality of blocks, the contour points corresponding to the contour point data. A block detecting means for obtaining a block in which a character exists, and a sentence as a feature quantity of the character image data by integrating a feature quantity count value of the block obtained by the block detecting means. An image processing apparatus comprising: character image data feature amount extraction means;