JP2507377B2 - Character recognition device using facsimile - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (1) Description of Field of the Invention The present invention relates to a character recognition apparatus using a facsimile, which recognizes character information in sheet information transmitted by a facsimile.

(2) Description of the prior art Conventionally, in the recognition of character information in sheet information transmitted by an encoded facsimile signal, scanning is performed for the number of scanning lines of code error within one page or code error within a specified number of scanning lines. The number of lines is counted, and if this value exceeds the specified value, the image signal is abnormal and the recognition process is stopped. For this reason, the total number of code errors per page or within the specified number of scan lines exceeds the permissible value, even if it can be determined that the code error scanning line does not exist in the character portion or if it exists in the character portion and the recognition is not affected. If so, the recognition process is stopped as an abnormal image signal. Therefore, the sender of the sheet cannot receive recognition processing even though there is no code error in the character part, or the number of code error in the character part is small but continuous, or code errors that greatly affect the recognition. Even in the case of type and the like, one error is counted as one time, and the total number of code errors is less than the allowable value, so that it may be erroneously recognized.

(3) Object of the invention It is an object of the present invention that even if there is a code error in the sheet information, if it is not a character part, it is not regarded as an image signal abnormality and is processed as a normal image signal and there is a code error in the character part. Even if it can be judged that the recognition rate is not affected by using the error scanning line position, the continuity of the position, and the total number of errors, normal processing is performed, and if not, it is recognized as an abnormal image signal. The purpose is to provide a means for accurately determining the normality of character information transmitted from a remote place such as a facsimile.

(4) Configuration of the Invention (4-1) Differences between Features of the Invention and Prior Art The present invention examines a code error existing only in a character portion when sheet information transmitted by a facsimile signal is decoded,
Check the number of consecutive code errors in the character part and the distance from one code error to the next code error, and check whether the image signal corresponding to the character part is normal with the count value obtained by weighting the code error for each code error state. The main feature is that it is possible to determine whether or not. It differs greatly from the conventional technique in the accuracy of the recognition result.

(4-2) Example FIG. 2 is a diagram showing an example of a sheet used in the present invention, 1 is a sheet, 2 is a left character line designation mark, 3 is a right character line designation mark, 4 Is a character entry frame and a character line designation mark
It is assumed that 2, 3 are black, the character entry frame 4 is at a predetermined position from the character line designation marks 2, 3, and is described in dropout color.

3 (A), (B), (C), (D), and (E) are Mo.
It is a figure which showed an example of the decoding processing system of a dified Huffman (MH) code | symbol, and n1-n5 are the original image signals of each decoded scanning line. Next, the code error processing will be described with reference to FIG.
First, when the MH code is normally decoded, the original image signal as shown in FIG. 3 (A) is obtained. When there is an error in the run length code corresponding to the scanning line n3 in the MH code, the original image signal of the scanning line n3 is replaced with n2 (previous line replacement), and the original image signal as shown in FIG. 3B is obtained. Scan line
If the End of Line (EOL) code indicating the end of n3 has an error, the scanning lines n3 and n4 are replaced by the original image signal of the scanning line n2 (line loss), and the original image signal as shown in FIG. Becomes In addition, the decoding result when there is an error in the length code corresponding to the scanning lines n3 and n4 in the MH code is shown in FIG.
FIG. 3E shows the decoding result when the run length code corresponding to the scanning lines n3 and n5 in the MH code has an error. Code errors are generally handled in this way.

Fig. 4 is a diagram showing an example of a character line detection method.
Is the upper scanning line of the character line designation mark 2, RT is the upper scanning line of the character line designation mark 3, LB is the lower scanning line of the character line designation mark 2, and RB is the lower scanning line of the character line designation mark 3. Is. Next, the character line detection method will be described with reference to FIG. However, it is assumed that the sheet 1 has been input by a facsimile in a left-up direction, and the character line designation marks 2 and 3 can be read with a width of W pixels. First, the original image signal in the vicinity of the character line designation marks 2 and 3 on each scanning line is examined. Then, the upper end of the character row designation mark 2 is determined by first detecting the scanning line LT in which the black pixels consecutive in W pixels are present at the left end, and the scanning line RT in which the black pixels consecutive in W pixels are present at the right side first. Is detected, the upper end of the character line designation mark 3 is determined. Next, the lower end of the character line designation mark 2 is determined by detecting that the scanning line LB having W pixels consecutive black pixels at the left end and the absence of W pixels consecutive black pixels at the next scanning line are detected, and the The lower end of the character line designation mark 3 is determined by detecting that there is no black pixel in which W pixels are continuous in the scanning line RB having a continuous black pixel and the next scanning line. However, the value of the W pixel is determined within the range of the cotton density of the facsimile × the width of the character line designation marks 2 and 3 in consideration of the reading error of the facsimile.

FIG. 1 is a block diagram of an embodiment of the present invention. 10 is an interface circuit for performing a facsimile procedure, 11 is a decoding circuit for decoding an MH code, and 12 is an original image signal sent from the decoding circuit 11. A character line detection circuit that detects the character line designation marks 2 and 3, and 13 stores the scanning line position and type of the code error detected during decoding of the MH code by the decoding circuit 11 and the character detected by the character line detection circuit 12. An error memory judging circuit for judging whether or not the character can be normally recognized based on the position information at the beginning and the end of the line, and 14 is whether the character line from the error memory judging circuit 1 can be normally recognized or abnormal. In response to the notification, the character recognition circuit that recognizes the original image signal of the character line sent from the character line detection circuit 12 only when normal, 15 is the recognition result and error memory determination circuit sent from the character recognition circuit 14. Image signal abnormality notification from 13 This is an interface circuit for transmitting to a device or the like.

Next, FIG. 1 will be described. First, the interface circuit
When 10 receives the facsimile signal, it extracts the MH encoded image signal and sends it to the decoding circuit 11. Decoding circuit 11 receives MH
The encoded image signal is decoded and the original image signal is transmitted to the character line detection circuit 12. At this time, if a code error is detected, decoding processing is performed by the method described with reference to FIG. 3 and the sub-scanning position of the error scanning line is notified to the error memory determination circuit 13. Character line detection circuit
12 detects the upper end scanning lines LT and RT and the lower end scanning lines LB and RB of the character line designation marks 2 and 3 in the sub scanning direction by the method shown in FIG.
This value is notified to the error memory determination circuit 13 and the scanning line
The image signal in the range of the scanning line RB from LT is transmitted to the character recognition circuit 14. The error memory judgment circuit 13 scans the scanning line LT based on the value notified from the character line detection circuit 12 and the value notified from the decoding circuit 11.
Code error between the scan line and scan line RB and the distance from the code error to the next code error are checked, and weighting is performed for each code error state to count (for example, the third
As shown in FIG. 3B, the code error of one scanning line having no code error more than two scanning lines before and after is 1, and when the code error is detected continuously for two scanning lines as shown in FIG. 3D. The error 1 scan line is counted as 3, and the error 1 scan line is counted as 2 when a code error every other scan line is detected as shown in FIG. 3 (E)). When the count value of the error scanning line of the character line exceeds the allowable value, the character recognition circuit 14 and the interface circuit 15 are notified that the image signal of the character is abnormal, and otherwise. The character recognition circuit 14 is notified of recognizability. Character recognition circuit
The reference numeral 14 discards the image signal of the character line when the error memory determination circuit 13 notifies that the image signal of the character line is abnormal, and the error memory determination circuit 13 can recognize the character line. When the notification indicating that is received, the character position is detected and recognized from the character line designation marks 2 and 3, and the recognition result is transmitted to the interface circuit 15. The interface circuit 15 is a character recognition circuit
The recognition result received from 14 is transmitted to an external information processing device or the like. However, when the interface circuit 15 receives a notice from the error memory determination circuit 13 that the image signal is abnormal, it notifies the external information processing device or the like of the image signal abnormality.

In the present embodiment, in the decoded image signal, the number of consecutive code errors in a character line and the distance from the code error scanning line to the error scanning line are checked next, and weighting is performed for each error state to count the characters. Although the image signal abnormality in the line was determined, it can also be determined by counting the total number of errors in a character line by setting the weights of the code errors in one scanning line to 1, and if there is a type of code error, weight the code error type. It is also possible to judge by counting the code error in the character line with the value obtained by performing the above (for example, one code error having a great influence on recognition is 2, and usually 1). It is also possible to make a judgment by counting these in combination. Further, in the present embodiment, the method of detecting the character line by detecting the character line designation mark is shown, but the character line can be detected by projecting the character portion instead of this method.

Due to this structure, it is considered that there are many errors in the character part that affect the recognition, those that continuously affect the recognition result even if there are few code errors in the character part, etc. Along with processing, it can be recognized as a normal image signal other than these.

As is clear from this result, it was possible to greatly improve the accuracy of the recognition result as compared with the conventional technology.

(5) Invention of Effect As described above, according to the present invention, the distance of the next code error is checked from the number of consecutive code errors appearing in the character portion of the sheet and the code error, and weighting is performed for each state. In order to judge whether normal recognition is possible with the counted value of the value,
If it is other than the character part, normal processing can be performed even if there are many code errors, and if there is an error state that affects recognition even if there are few code errors in the character part, it can be determined that the image signal is abnormal.
Therefore, there is an advantage that more accurate recognition can be ensured than the method of judging the image signal abnormality by simply using the total number of code errors of one sheet or the total number of code errors within the specified number of scanning lines.

[Brief description of drawings]

FIG. 1 is a block diagram of an embodiment of the present invention, FIG. 2 is a diagram showing an example of a sheet used in the present invention, FIG. 3 is a diagram showing an example of a decoding processing system of MH code, and FIG. The figure shows a diagram showing an example of a character line detection method. 1 …… Sheet, 2 …… Left character line designation mark, 3 …… Right character line designation mark, 4 …… Character entry frame, n1 to n5 …… Decoded original image signal of each scanning line, LT …… Left character line designation mark 2
Of the upper edge of RT, RT ... the scanning line that detected the upper edge of the right character line designation mark 3, LB ... the scanning line that detected the lower edge of the left character line designation mark 2, RB ... the right character Scan line that detected the lower end of line designation mark 3, 10, 15 ... Interface circuit, 11 ... Decoding circuit, 12 ... Character line detection circuit, 13 ...
Error memory judgment circuit, 14 …… Character recognition circuit.

Claims (1)

(57) [Claims] 1. A character recognition device for transmitting and recognizing a sheet in which character information is input by a facsimile, and decoding the encoded image signal of the sheet transferred from the facsimile to reproduce an original image signal. First means, second means for determining the error position when a code error occurs in the first means, and third means for detecting the position of the character portion in the sheet from the original image signal, The positional information of the character portion detected by the third means and the second information
Fourth means for checking the number of consecutive code errors in the character part and the distance from the code error to the next code error using the position information of the code error determined by the means, and the code error checked by the fourth means A code error is counted by a value weighted corresponding to each of the states, and the recognizability of the character portion is judged by judging the counted value.
And a character recognition device using a facsimile.