JPH0434793B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to an optical character reading device, and particularly to detection of displacement points required when extracting the outline of a character.

Optical character reading devices are called OCR and are widely used in data processing systems that handle large amounts of data.

[Conventional technology]

Conventionally, when reading data using OCR, the physical information of the form on which the data is written (dimensions, weight, etc.) and information such as logical data processing conditions are set in advance as format information, and this format Character recognition processing is performed according to the information.

FIG. 3 is a diagram showing an example of a conventional character reading device.

In the figure, NCNT is the main control unit, SCAN is the scanning unit,
MEMa is image memory, DISPLAY is display unit,
KEY is the keystroke part, RECOG is the recognition part, DIC is the recognition dictionary, MEMb is the extraction memory, MEMc is the format definition body, F-CNT is the format control part,
OUT is an external recording medium for output.

The recorded information written on the form SHEET is scanned by the scanning unit.
Read by SCAN through optical sensor, 2
The digitized result is stored in the image memory MEMa. The main control unit MCNT sends read information such as position information of fields on the format definition body MEMc to the recognition unit according to a predetermined procedure.
Send to RECOG.

The recognition unit RECOG extracts the corresponding character image from the image memory MEMa, stores it in the memory MEMb, extracts the characteristics of the character pattern, and converts it into a recognition dictionary.
Check with DIC.

Extraction memory MEMb of character reading device like this
It is necessary to extract displacement points in the horizontal and vertical directions as preprocessing when extracting the characteristics of the character pattern from the information transferred to the character pattern.

FIG. 4 is a diagram showing an example of a method for detecting a displacement point of information stored in the extraction memory MEMb, and shows a case where the number "2" is stored.

The information stored in the cutout memory MEMb is stored in word units; for example, in the first column in Figure 4, the 0 position is white, the 1 to 13 positions are black, and the 14 and 15 positions are white, so it is 16 bit information. , one word consists of 8 bits, one column consists of 2 words, and word (i ₁ , j ₁ ) is 8 bits “01111111”.
Similarly, the word (i ₁ , j ₂ ) is “11111100”
becomes. Processing in the recognition unit RECOG is performed in units of words.

In the figure, ○ marks indicate displacement points in the left-right direction, and × marks indicate displacement points in the vertical direction.

If you scan in the X direction from the top left, column 17 is all white, column 16 has a white → black displacement point ○ mark at position 5, and 12
There is a black → white displacement point ○ mark at the position, and the effective number 2 indicates that there are 2 displacement points in this 16 column. Also, in the 10th column, there are displacement points ○ marks at positions 13 and 15, and the effective number is 2.

Similarly, when scanning in the Y direction from the bottom left, in one line, there is a white-to-black displacement point × mark at position 1, a black-to-white displacement point × mark at position 4, and the effective number is 2, 14
In the row, there is a white to black displacement point x mark at position 8, and a black to white displacement point x mark at position 14, and the effective number is 2.

As described above, the conventional displacement point detection method performs two scans, one in the left-right direction and the other in the vertical direction, to determine the displacement points, respectively. Furthermore, in the case of detecting displacement points in the left-right direction, there is a drawback that the displacement points cannot be determined at the rightmost and leftmost points in a word without referring to information from other words adjacent to the right and left, respectively.

[Problem that the invention seeks to solve]

An object of the present invention is to eliminate the above-mentioned conventional drawbacks, to simultaneously detect displacement points in the horizontal and vertical directions, and to achieve high-speed displacement point detection by detecting displacement points in the horizontal direction only using the word concerned. It is.

[Means for solving problems]

The means for solving the problem include a means for binarizing the image information on the form and storing it in a memory, a means for cutting out the character part from the image information, and a means for cutting out the character image information or the character image. In a character reading device comprising means for comparing feature information obtained by extracting features from information with a standard pattern, and means for determining and outputting a reading result based on the matching result and pre-given format information, etc., Vertical displacement point detection required to extract the outline of the cut out character image consisting of multiple words uses information about the word (i, j) and the word (i-1, Take the exclusive OR with the information of j), and detect the displacement point in the left and right direction using the corresponding word (i, j)
Exclusive OR is performed between the information of the word (i, j) and the information obtained by shifting the information of the word (i, j) one bit to the right, and the displacement point of the leftmost bit of the word (i, j) is detected based on the word (i, j). This is achieved by a character reading device characterized in that this is performed by referring to the state provisionally registered by the rightmost bit of the word (i, j-1) to the left of the word (i, j-1).

[Effect]

According to the present invention, displacement points in the vertical direction are detected by performing an exclusive OR between the word in question and the word one line below it, and displacement points in the horizontal direction are detected by shifting the word one place to the right. After that, the exclusive OR is taken as a displacement point, and this can be processed in one phase by performing some additional processing, which has the great effect of greatly shortening the processing time compared to the conventional two-phase processing method. is born.

〔Example〕

FIG. 1 is a diagram illustrating an embodiment of a displacement point detection method for a character reading device according to the present invention.

FIG. 1a is a diagram showing a method for detecting displacement points in the vertical direction.

As mentioned above, the storage information in the extraction memory MEMb is a plurality of binarized data of read characters stored in the form of words in which each word consists of 8 bits, and the recognition unit RECOG stores the following binary data. Processing takes place.

In the figure, the word (i, j) indicates an 8-bit word (i, j) located in row i and column j. For example, of the 16 bits in any row in Fig. 4, the left half 8 bits or the right half 8 bits , indicating that is black-and-white information.

In addition, indicates the word (i-1, j) located in the i-1 row and j column, indicating that it is located directly below the word (i, j), and the black-white, black, black, black, black, and white information. have.

Here, the displacement point in the vertical direction can be found by taking the exclusive OR (EOR) of the information on the word (i, j) of and the information on the word (i-1, j) located directly below it. become.

The result is that the 2nd and 7th bits are "1" as shown in Figure 4, and this is the vertical displacement point (4th (X mark in the figure).

Next, FIG. 1b is a diagram showing a method for detecting displacement points in the left-right direction.

In this case, the information in the figure is the same as in Figure 1a, and when scanning the horizontal displacement point from left to right, the 3rd bit is "0" as is clear from the figure.
This is the transition point from "1" to "1", and the 7th bit is the transition point from "1" to "0".

To detect this displacement point, first, the information of word (i, j) of is shifted one bit to the right to obtain information.

Next, information is obtained by taking the EOR of the information, and the location of "1" is determined as the displacement point in the left-right direction.

This shows that this is the horizontal displacement point (marked with a circle in FIG. 4) at the word position in row i and column j when scanning in the X direction from the lower left in FIG.

However, in the above method, it is impossible to detect whether or not the leftmost bit of word (i, j) is a displacement point unless there is a bit further to the left.

Therefore, the corresponding word (i, j) must be detected by comparing it with the word (i, j-1) to the left. However, if the corresponding word (i, j) is the leftmost word, the leftmost bit is not regarded as a displacement point. The information on the left adjacent word (i, j−
1), and when the rightmost bit is "1" (black), the leftmost bit of the corresponding word (i, j) on the right is temporarily registered as a displacement point.

If the leftmost bit of the corresponding word (i, j) is "1" (black) as shown by the information in , there is no change from the rightmost bit of "1" (black), and it is not a displacement point. Invalidate the temporary registration and reduce the number of displacement points by one.

If the leftmost bit of the corresponding word (i, j) is “0” (white), the rightmost bit of the word (i, j) is “1”.
(black), and the temporarily registered displacement point is confirmed.

Additionally, if the rightmost bit of the word (i, j-1) on the left is “0” (white), the leftmost bit of the corresponding word (i, j) on the right is provisionally registered as not a displacement point. In this state, if the leftmost bit of the word (i, j) is "0" (white), it is confirmed that there is no change, and if it is "1" (black), it is registered as a displacement point.

In this way, it is possible to detect whether the leftmost bit of the word (i, j) is a displacement point if the rightmost bit of the word (i, j-1) to the left is "1". If it is registered, and it is "0", it is not provisionally registered, and it can be detected by referring to the state of this provisional registration.

Thereby, the displacement point can be detected using only the word without referring to the leftmost word or the right-most word.

The procedure for detecting a displacement point using the above processing will be explained below.

FIG. 2 is a diagram showing a procedure for detecting a displacement point according to the present invention for the number "2".

The processing procedure consists of process a going from the bottom left to the top, process b going back to the bottom right, and process c going up. Process a clears the word (i ₁ -1, j ₁ ) to 0.

The vertical displacement point detection process shown in FIG. 1a is performed. That is, word (i ₁ , j ₁ ) and word (i ₁ −
1, j) and detect the vertical displacement point of the word position in row _i and column _j .

Detection of displacement points in the left and right direction shown in FIG. 1b is performed. In other words, perform FOR between the word (i ₁ , j ₁ ) and the information shifted one bit to the right, i ₁ line,
j Perform horizontal displacement point detection processing for the word position in _one column.

Increment i row by 1.

The processes of ) to ) are repeated until the i row becomes the i _H+1 row, and the process a is completed.

Process b changes row i to row i ₁ and changes column j ₁ to column j ₂ .

Process c, like process a, repeats the process until row i becomes row i _H+1 . In this case, the leftmost bit of column _j2 is detected by referring to whether or not it has been provisionally registered by the rightmost bit of column _j1 , as described above.

With the above steps, detection of the displacement point of number "2" in FIG. 2 is completed.

In the displacement point detection process according to the present invention, the two scans in the vertical and horizontal directions, which were required in the conventional process, can be performed in the same phase, so the process can be performed approximately 30% faster than the conventional method.

〔Effect of the invention〕

As described in detail above, according to the present invention, displacement point detection processing in the vertical direction and horizontal direction can be performed in the same phase, and in the horizontal direction displacement point detection processing, only the word (left or right) can be detected. This method has the great effect of increasing the processing speed compared to the conventional method since it is possible to detect displacement points (without referring to words).

[Brief explanation of drawings]

FIG. 1a is a diagram showing a method for detecting displacement points in the vertical direction. FIG. 1b is a diagram showing a method for detecting displacement points in the left-right direction. FIG. 2 is a summary diagram of the displacement point detection procedure according to the present invention. FIG. 3 is a diagram showing an example of a conventional character reading device. FIG. 4 is a diagram showing an example of information stored in the extraction memory MEMb. In the figure, NCNT is the main control unit, SCAN is the scanning unit,
MEMa is image memory, DISPLAY is display unit,
KEY is the keystroke part, RECOG is the recognition part, PIC is the recognition dictionary, MEMb is the extraction memory, MEMc is the format definition body, F-CNT is the format control part,
OUT is an external recording medium for output.

Claims (1)

[Scope of Claims] 1. A means for binarizing image information on a form and storing it in a memory, a means for cutting out a character part from the image information, and a feature of the cut out character image information or the character image information. In a character reading device that is equipped with a means for comparing characteristic information obtained by extraction with a standard pattern, and a means for determining and outputting a reading result based on the matching result and pre-given format information, etc., Detection of displacement points in the vertical direction, which is necessary to extract the outline of the cut out character image consisting of Perform exclusive OR with the information, and detect the displacement point in the left and right direction using the corresponding word (i, j)
Exclusive OR is performed between the information of the word (i, j) and the information obtained by shifting the information of the word (i, j) one bit to the right, and the displacement point of the leftmost bit of the word (i, j) is detected based on the word (i, j). A character reading device characterized in that the character reading is performed by referring to a state provisionally registered by the rightmost bit of the word (i, j-1) to the left of the word (i, j-1).