JPS6239459B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a noise image removing device used in a preprocessing section of an optical character reading device to perform character separation (cutting) processing.

In optical character reading devices, it is necessary to perform preprocessing to separate (cut out) written characters one by one. If each character is insufficiently separated (cut out) in this preprocessing and unnecessary parts are mixed into the separated characters to a certain extent, it becomes impossible to identify the characters. This invention solves the conventional difficulties in the preprocessing section of the optical character reading device mentioned above,
This makes it possible to remove unnecessary image portions from an image signal representing an image within a character frame.

For this reason, in the noise image removing apparatus of the present invention, the out-of-frame image detection means detects an image existing between adjacent character frames as an out-of-frame image. The unnecessary image detection means detects whether an image exists within the character frame following the detected image outside the frame, and if an image exists, the image existing within the character frame is made unnecessary based on the image outside the frame. Extract as an image. The unnecessary image removing means uses the extracted unnecessary image to remove a corresponding unnecessary image portion from the image signal representing the image within the character frame to obtain a noise-free image signal.

According to the noise image removal device of the present invention, for example, when a character as shown in FIG. Remove as shown. Characters written on forms, etc., which are the subject of this invention, are scanned by an optical scanner, etc., and each pixel is quantized into binary values of 1 and 0, and arranged in a vertical and horizontal manner as shown in Figure 1. The resulting image is made up of pixels. An image signal based on this image is stored in the storage circuit 1 of the noise image removing apparatus of the present invention shown in FIG. For ease of explanation, each character quantized in this way is represented by one vertical pixel line, that is, one line is 16 bits, and each character is represented by a character area of 12 lines and a boundary area of one line. do.

FIG. 1 shows a case where the first character 11 is embedded in the second character frame 12, and the second character 13 is embedded in the first character frame 14. Also, it is assumed that addresses n and n+13 among the addresses indicating lines in the figure are outside the character frame calculated and set in advance by the character entry pitch and the like. In other words, address n is character frame 12
The outside of the frame between and 14 is shown.

Generally, 16 bits per line is often insufficient due to variations in character size and character position, and several words per line are often required. In addition, although there are cases in which it is actually impossible to completely separate characters simply by the input pitch, this can be solved by extending and applying this invention in more detail to areas outside the frame.
For the sake of simplicity, the explanation here will be based on the assumption that characters can be completely separated depending on the input pitch.

As shown in Figure 2, if these two characters are separated at the boundary between the first character 11 and the second character 13, unnecessary parts will appear in the first and second characters.
A pattern including unnecessary images with N ₁ and N ₂ remaining is obtained. Characters that include such unnecessary image portions that become noise generally become unidentifiable. According to the noise image removing device of the present invention, in this case, 1
The inverted L-shaped unnecessary image part _N1 included in the character pattern of the first character and the bar-shaped unnecessary image part _N2 included in the second character pattern are detected as noise images, and these are divided into 1 and ₁ in Fig. 3. A pattern with these removed is obtained as shown in ₂ . That is, in the noise image removal device of the present invention, if the entire pattern of the first character including the unnecessary image portion _N1 is L, and the pattern of only this unnecessary image portion is _N1 , then the image signal of the first character is L.・A pattern as shown in FIG. 3a is obtained in which the unnecessary portion _N1 is removed by performing a logical operation of ₁ .

First, a method for extracting the shape of an unnecessary portion included in a pattern will be described below.

First of all, in this invention, the unnecessary image portions that become noise images are: (a) The unnecessary image portion is separated from the characters, that is, the unnecessary image portion is not in contact with the characters, and (b) The unnecessary image portion is separated from the characters by the m line ( It is defined as satisfying two conditions: (for example, 4 lines) or more and not extending beyond the character boundary.

An algorithm for removing this unnecessary image portion will be explained below.

First, the image signal between the character frames 14 and 2 character frame 12 in Figure 1 is detected (the contents of address n in the memory), and as shown in Figure 4 a, the image signal corresponding to the two black blocks is detected. A pixel with a logical value of "1" is obtained as an out-of-frame image signal. If the content between the character frames is not "0", there is a possibility that an unnecessary image part that becomes a noise image exists, and if it is "0", it means that no unnecessary part exists. In the case of FIG. 4a, since two black blocks are detected between the character frames, there is a possibility that there are two unnecessary image parts that become noise images.

The detected out-of-frame image signals of the two black blocks shown in Fig. 4a are separated into respective blocks as shown in Fig. 4b and c.
Unnecessary image parts are determined based on n ₁ and n ₂ .
Since the operation at this stage can be easily realized by performing a simple shift operation, a detailed explanation of the operation will be omitted.

First, starting from the position of the out-of-frame image signal _n1 in FIG. 4b, images continuous with this are tracked to check whether they are unnecessary parts or not. Therefore, the black pixel corresponding to the out-of-frame image signal _n1 shown in FIG. 4b is read out from the storage circuit 1 in FIG. 7 showing the configuration of an embodiment of the present invention as shown in FIG. Set in register 5. In this invention, it is searched whether or not there is a black pixel within the character frame that is continuous with this outside-frame image signal _n1 . A search area is set in advance for this search. That is, the arithmetic operation of the arithmetic circuit 3 is selected to be in the upward shift state, and the value set in the register 5 is shifted one bit higher on the same ground line.
The output is set in register 6.

Next, the arithmetic operation of the arithmetic circuit 3 is selected as OR, and the values of the registers 5 and 6 are ORed. The result of this OR operation is given to register 6 and stored.

Arithmetic circuit 3 is then selected to shift down,
The value set in register 5 is shifted one bit lower on the same first line, and its output is set in register 5. The value set in the register 5 and the value obtained in the previous operation and stored in the register 6 are selected for the arithmetic operation of the arithmetic circuit 3 as a logical sum. This calculated value is set in register 5. This result corresponds to a black pixel image as shown in FIG. 5b. As described above, in the present invention, the detected out-of-frame image is set in the register 4, and a search area including the out-of-frame image and one pixel each above and below the out-of-frame image is set by calculation based on the out-of-frame image.

When the search area is set, the first vertical pixel line adjacent to the out-of-frame image is read out, and the first vertical pixel line corresponding to the search area is read out.
It is determined whether a black pixel exists for each pixel in the vertical pixel line. To do this, first, the contents of the black pixel image shown in FIG. It is set to 6. The arithmetic operation of the arithmetic circuit 3 is selected as AND, and the contents of the black pixel image shown in FIG. A logical AND operation with the contents of is performed. The result of this operation is set in register 5, which corresponds to the black pixel image shown in FIG. 5d.

In this way, if the address between character frames is subtracted and the operation is performed with the black pixel image at address n-1 for the first character, there is a possibility that unnecessary parts are inserted from the right side. Because there is. In the case of the second character, if there is a possibility that an unnecessary part is digging in from the left side, add the address of the memory circuit 1 as the image outside the frame and the image of the first vertical pixel line for the image outside the frame, and write the black at address n+1. Perform a logical AND operation with the pixel image, and if there is a possibility that unnecessary parts are inserted from the right side even in the case of the second character, subtract the address in the same way as the first character and use the black pixel at the address obtained. Performs an AND operation with the image.

In this example, if the value of the logical product of the black pixel image created based on the image outside the frame detected between the frames of two characters and the black pixel image at address n-1 obtained by subtracting the address is 0, then The unnecessary image portion _N1 is separated from the first character (2 in this case). However, in this embodiment, since this value is not 0 as shown in FIG. 5d, the shape of the unnecessary image portion _N1 is further tracked. This unnecessary image portion may be embedded within the character frame by one or more lines. In order to clarify this state, a noise gate signal corresponding to each line is generated. The noise gate signal in this case is equal to the value obtained by extracting the shape itself of each line of the unnecessary portion. For example, the noise gate signal corresponding to the unnecessary image portion included in the first character is as shown in FIG. 6a.

First, it is tracked whether or not the unnecessary image portion _N1 extends upward from the black pixel image shown in FIG. 5d. That is, first, the arithmetic circuit 3 is selected to be in the upward shift state and a shift operation is performed on the black pixel image shown in FIG. 5D that has been set in the register 5 in the previous operation. This result is set in register 6. Arithmetic circuit 3 is selected for a logical sum operation, and the values of registers 5 and 6 are logically summed.
As a result, an image corresponding to the black pixel image shown in FIG. 5e is obtained. This result is set in register 5.

Then from the memory circuit 1 via the data bus 7
The contents of the black pixel image shown in FIG.
It is calculated by The result of this AND operation is shown in Figure 5g.
An image corresponding to the black pixel image shown in is obtained.

Next, the arithmetic circuit 3 is selected to be in the exclusive OR state, and the exclusive logic between the value obtained by the above AND operation and set in the register 6 and the contents of the black pixel image shown in FIG. 5e in the register 5 is performed. A sum operation is performed. As a result, an image corresponding to a black pixel image as shown in FIG. 5h is obtained.

The fact that the result of the exclusive OR shown in h of the figure is 0 means that the unnecessary image part of the line at address n-1 extends upwards, and the fact that the result of this exclusive OR is not 0 means that the result of the exclusive OR is not 0. This means that the unnecessary image portion of the line at address -1 does not extend further upwards. Therefore, the upper limit value of the black pixel image shown in FIG. 5D is repeatedly shifted on the same ground line until the value of the exclusive OR shown in FIG. The logical product of the image and the black pixel image at address n-1 is repeated to extract the position of the most significant bit of the unnecessary image portion on the 12th line corresponding to address n-1. In this example, 1
After this operation, the exclusive OR value shown in Figure 5h is no longer 0, so this operation ends and the black pixel image obtained initially, shown in Figure 5D, becomes part of the noise gate signal. It can be seen that the unnecessary image portion _N1 does not extend upward from the starting point d in FIG. 5 on the line at address n-1.

Next, since there is a possibility that the unnecessary image portion of the 12th line corresponding to address n-1 extends downward, it is necessary to perform further downward tracking to confirm this. The same line lower limit value of the black pixel image shown in Fig. 5d, which was found to form part of the noise gate signal by the above-mentioned tracking, is shifted down by 1 bit, and the black pixel image shown in Fig. 5d. By calculating the logical sum with the pixel image, a black pixel image as shown in FIG. The black pixel image thus obtained and the contents of the black pixel image at address n-1 in the storage circuit 1 are logically ANDed to obtain the black pixel image shown in FIG. As in the case of upward tracking, if we take the exclusive OR of the black pixel image shown in FIG. Become. This state is shown in figure m.

This state means that the unnecessary image portion N1 on the 12th line corresponding to address n- ₁ extends further down. Therefore, in this case, downward tracking is continued in the same manner as in the case of upward tracking. That is, the fifth
The black pixel image obtained by shifting the lower limit value on the same ground line of the black pixel image shown in FIG. Performs a logical product with the black pixel image at address -1. Next, an exclusive OR is calculated between the black pixel image obtained by this AND operation and the black pixel image obtained by sequentially extending the lower limit pixels downward, and this operation is performed until this value is no longer 0. By repeating such calculations, the noise gate signal shown in FIG. 5n is finally obtained as the noise gate signal of the 12th line corresponding to address n- ₁ , which is the first vertical pixel line for the out-of-frame image n1.

The black pixel portion shown in FIG. 5n obtained in this way is a black pixel portion that exists in the first vertical pixel line continuous with the out-of-frame image. The image shown in FIG.
is memorized. Next, it is determined whether or not there is a black pixel continuous with the black pixel portion shown in FIG. 5n in the second vertical pixel line adjacent to the first vertical pixel line. In this case as well, n-1 shown in FIG.
Based on the noise gate signal of the 12th line corresponding to the address, it is sufficient to associate this with the black pixel image of FIG. 5a and start the arithmetic operation. A detailed explanation of the calculation process in this case will be omitted, but ultimately the fifth
The noise gate signal shown in Figure o is obtained.

After the extraction of the 11th line corresponding to address n-2, which is the second vertical pixel line, is completed, the noise gate signal of the 10th line corresponding to address n-3, which is the third vertical pixel line, is extracted. . In this case n-
Based on the noise gate signal of the 11th line corresponding to address n-2 shown in o in the same figure, which was obtained by extracting the 11th line corresponding to address 2, this is associated with the black pixel image in a in the figure It is a good idea to start arithmetic operations in the same way. In this case as well, a detailed explanation of the process of arithmetic operations will be omitted, but since the arithmetic output becomes 0 at the stage corresponding to the operations shown in Figure 5 b and c, it is important to note that the unnecessary part _N1 does not embed more than 3 lines into the character part. become.

In this embodiment, the operation of creating the noise gate signal based on the black pixel _n1 in FIG. 4b is completed in the process of the operations up to this point, and as a result, the noise gate signal as shown in FIG. 6a is generated. This is the memory circuit 1
is memorized.

Next, a noise gate signal is generated in the same manner as in the case where the position of black pixel n ₂ is used as the starting point and the position of black pixel n ₁ is used as the starting point, as shown in FIG. 4c. A detailed explanation of the operation in this case will be omitted, but in this case, the above-mentioned condition b of the unnecessary part that does not extend beyond the character boundary by more than m lines (for example, 4 lines) is not satisfied, and the operation is performed by more than m lines. Since the noise gates for lines 12 to 9 cannot be separated even after doing this, the noise gates for lines 12 to 9 are reset and these are not determined to be unnecessary image parts.

Through the series of operations described above, a noise gate signal for the first character as shown in Figure 6a is obtained, and this noise gate signal is completely identical to the shape of the unnecessary image part shown in Figure 2a. matches.

Therefore, by calculating the logical product of the character pattern existing in the character frame including the unnecessary image part shown in FIG. 2a and the complement of the signal of the unnecessary image part of the first character shown in FIG. 6a, It is possible to remove only the unnecessary image portion that becomes a noise image, and the image signal shown in FIG. 3a is obtained.

Processing for the second character is performed in the same way, and in this case, starting from the out-of-frame images n ₁ , n ₂ , these out-of-frame images n ₁ ,
_n2 and the address of the storage circuit 1 as the image of the first vertical pixel line adjacent thereto may be added to perform an operation with the black pixel image at address n+1.
By performing such processing, the result shown in Fig. 6b is
The noise gate signal shown in is obtained. Therefore, the second
The logical product of the character pattern existing in the character frame of the second character including the unnecessary image part _N1 as shown in Figure b and the complement of the signal of the unnecessary image part of the second character shown in Figure 6b. By calculating , an image signal as shown in FIG. 3b from which unnecessary image portions have been removed can be obtained.

Furthermore, regarding the third character shown in Figure 1, the pixel signal value (n+
Since the value at address 13) is 0, it is determined that no unnecessary image portion exists and no noise image is generated. Furthermore, it goes without saying that even when a plurality of unnecessary image parts are mixed in one character, according to the present invention, the same process can be performed to remove the unnecessary image parts that are determined to be noise images. Note that the circuit of the embodiment of the present invention shown in FIG. 7 can generally be easily constructed using a microprocessor under current digital technology. The algorithm for removing unnecessary parts described above is a combination of logical AND, logical OR, exclusive OR, and shift operations, and such arithmetic operations can be easily realized using a microprocessor.

As described in detail above, the noise image removal device of the present invention has a simple configuration and operates by combining a simple shift operation, an OR operation, an AND operation, and an addition operation without requiring complex operations. It is possible to quickly and accurately obtain a character image from which noise images have been removed.

[Brief explanation of the drawing]

Figure 1 is a diagram showing quantized characters stored in a memory circuit, Figure 2 is a diagram showing each character separated by a calculation boundary line, and Figure 3 is a diagram showing a state in which each character is separated by a calculation boundary line.
The figure shows a state in which unnecessary image parts have been removed from the pattern in Fig. 2, and Fig. 4 shows a state in which the image existing between the character frames of the first and second characters has been divided into two black blocks. Figure 5 is a diagram showing the calculation results in the process of removing the unnecessary image part of the first character, Figure 6 is a diagram showing the extracted unnecessary image part of the first and second characters, and Figure 7 1 is a block diagram showing the configuration of an embodiment of the present invention. FIG. 1: Memory circuit, 2: Address register, 3: Arithmetic circuit, 4, 5, 6: Register, 7: Data bus.

Claims (1)

[Claims] 1. Means for converting a target image consisting of pixels arranged vertically and horizontally into a binary image in which each pixel takes one of two values, and converting a binary image existing between adjacent character frames into an outside-frame image. an out-of-frame image detection means for detecting an out-of-frame image; It is determined whether or not there are continuous black pixels, and the determined black pixel portion is expanded vertically to form a search area, and each corresponding pixel of this and the second vertical pixel line is continuous to the black pixel portion. It is determined whether a black pixel exists or not, and the same process is repeated sequentially for the vertical pixel lines in the character frame, and the black pixels of that vertical pixel line are added to all pixels corresponding to the newly set search area. If it is determined that the black pixels do not exist, the black pixel portion that has been determined to be continuous up to that point is used as a noise gate signal. noise gate signal detection means that does not use a black pixel portion that is determined to be continuous with the image outside the frame as a noise gate signal related to the image outside the frame; A noise image removal device comprising: means for removing a corresponding portion;