JP2578959B2 - Character / photo region determination device, image input device using the same, and optical character recognition device - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a character / photograph area determination technique, an image input technique using the same, and an optical character recognition technique, and more particularly, to a character area and a photograph area or a dense black area. The present invention relates to a technique suitable for processing to cut out an image having a complicated area or a character written on a non-white background.

[Conventional technology]

For example, in inputting an original image in a facsimile, a digital copying machine, an electronic file device, or the like, a character area in which the contrast with the background is relatively clear and a multi-tone image area such as a photograph are accurately distinguished. It is important to set appropriate thresholds and digitize them from the viewpoint of avoiding deterioration of the input image due to fixing the thresholds as one of the areas.

For this reason, the following several techniques have been conventionally proposed.

For example, the technology disclosed in Japanese Patent Application Laid-Open No. 61-146070 discloses an optimal binary value according to image characteristics by using a binarization circuit having a different threshold value in addition to a character binarization circuit for area determination. Is trying to make it possible.

Further, the technology disclosed in JP-A-61-169083 is
By identifying the edge portion (peripheral portion) of the line image based on the number of white or black pixels in the pixel adjacent to the target pixel in the binarized image, the image can be accurately and in real time formed with a small-scale circuit configuration. Trying to be identifiable.

Further, in the technique disclosed in Japanese Patent Application Laid-Open No. 61-80964, a character image is subjected to a spatial filter by applying a spatial filter to a pixel of interest and its surrounding pixels, and comparing the output with a predetermined threshold to provide a discrimination signal. An image information source in which a halftone image and a halftone image are mixed is intended to be accurately identified.

The spatial filter disclosed in Japanese Patent Application Laid-Open No. 61-80964 is a kind of two-dimensional information calculation for digital image information, and is generally called filtering.
A linear filter, one of such filtering processes, is described in detail in 3.2 Filtering on page 34 of “Industrial Image Processing” by Shoji Ejiri, published by Shokodo Co., Ltd., first edition on May 31, 1988. Have been.

[Problems to be solved by the invention]

However, in the technique disclosed in Japanese Patent Application Laid-Open No. 61-146070, when the white level of the background is near the threshold value for binarizing characters as in a catalog manuscript of color printing, or when the internal background of dense characters such as kanji characters is used. When the background modulation level is higher than the threshold value for character binarization because the degree of scanner modulation is not 100%, there is a problem that the background becomes difficult to detect.

According to the technique disclosed in Japanese Patent Laid-Open No. 61-169033, an edge portion of a line drawing is output as a character binary signal with respect to a pseudo halftone signal. Do not know how to improve misjudgment.

Further, the technology disclosed in Japanese Patent Application Laid-Open No. 61-80964 is disclosed in A / D
Although the two-dimensional information calculation for the multi-value output of the converter is performed, the hardware of the two-dimensional information calculation for the multi-value output tends to increase in proportion to the number of bits of the multi-value information. He does not recognize the suppression of the scale of a logic circuit when performing an operation.

Therefore, an object of the present invention is to provide a method for processing a character or a line drawing even when the white level of the background is close to the threshold for character binarization, or when the white level of the background is higher than the threshold for character binarization such as the internal background of a dense character. An object of the present invention is to provide a character / photograph area determination technology capable of detecting a background.

Another object of the present invention is to provide a method for performing edge enhancement processing as an alternative means or a reinforcing means for character / photograph area block determination, which is a problem in the edge around the black of a photograph or the false halftone signal conversion inside a character. It is an object of the present invention to provide a character / photograph area determination technique capable of reducing the number of determinations.

It is still another object of the present invention to provide a character / photo region determination technique capable of accurately determining a character / photo region.

Still another object of the present invention is a character / photograph area determination technology capable of realizing a two-dimensional information calculation processing method, which is one of the character / photograph area block determination processing methods, with a relatively simple circuit configuration. Is to provide.

Still another object of the present invention is to provide an image input technique capable of simplifying signal processing after image input.

Still another object of the present invention is to provide an optical character recognition technology capable of simplifying specifications of a form to be input.

The above and other objects and novel features of the present invention are as follows.
It will be apparent from the description of this specification and the accompanying drawings.

[Means for Solving the Problems] Of the inventions disclosed in the present application, the outline of a representative one will be briefly described as follows.

That is, the character / photo region determination apparatus according to the present invention detects a background by two-dimensional information calculation and outputs the binary background information, and a character / photo region by two-dimensional information calculation from the background information. Based on a region determination unit to be determined and a region determination result output from the region determination unit, a character binarized signal binarized with a certain threshold and coming from outside, and the threshold is changed for each pixel. Area switching means for selecting and outputting any one of the pseudo halftone signals binarized by the image processing apparatus, wherein the background detecting means has a higher threshold (darker density) than a character binarization circuit for generating a character binarization signal. ), A second binarization circuit for detecting dark black, and a second binarization circuit for background detection capable of switching between two thresholds, a lower threshold (lighter density) and a higher threshold than the character binarization circuit. Value A first two-dimensional information calculating means for performing a black density determination in units of n × n pixels on the basis of the output of the first binarization circuit, and an output of n × n pixels on the basis of the output of the second binarization circuit. A second two-dimensional information calculation unit that has two functions for performing white density determination as a unit and detects white in the background and outputs it as a white detection signal; The area determining means performs an operation of switching between the threshold value of the binarization circuit of 2 and the two functions of the second two-dimensional information calculating means, and when n and m are arbitrary positive integers, n × n pixel units A second two-dimensional information calculation unit having two functions for performing a determination operation with the number of different white blocks as a threshold value for each m × m block plane, using the background information output to Select one output of two functions of the fourth two-dimensional information calculation means A second switching means for outputting, and the output of the second switching means being m × m
A second two-dimensional information operation means having a plurality of operation functions capable of performing S (S is a positive integer) times of different two-dimensional information operations in two different ways on the block plane; The character / photo area determination results obtained from the area determination result, which is the final determination result after S times of the dimensional information calculation means, and the two threshold determination results of the fourth two-dimensional information calculation means are sequentially stored, and the current m The m × m block is obtained by using the character / photo area determination result in the × m block plane and the character / photo area determination result obtained from the area around the current m × m block plane. A global area determining means for globally determining a character / photograph area from a background wider than a plane, and a second area based on the determination result of the global area determining means.
And a plurality of calculation functions of the fifth two-dimensional information calculation means.

Further, the image input device of the present invention is an image input device using a black or gray background plate for detecting a form shape, wherein an image sensor for one-dimensionally reading image data from a target form, and an image sensor An analog-to-digital (AD) converter for digitizing an image signal read by a digital camera, a normalization circuit for performing a logarithmic conversion process, a shading process, and the like on the digitized form image signal to obtain a normalized signal; A binarization circuit for binarizing the signal and outputting it as a character binarization signal, a pattern generator for generating a specific identification pattern signal, and a character binary for receiving a normalized signal and generating a character binarization signal A first binarization circuit that detects dark black with a higher threshold (darker density) than the digitizing circuit, and two thresholds: a lower threshold (lighter density) and a higher threshold than the character binarization circuit A second binarizing circuit for detecting a background whose value can be switched, and a first two-dimensional information calculating means for performing a black density determination on an output of the first binarizing circuit in units of n × n pixels And a second two-dimensional information calculation that has two functions for performing white density determination on the output of the second binarization circuit in units of n × n pixels, and detects white of the background and outputs it as a white detection signal. Means, the first two
A background detection circuit for performing an operation of switching between a threshold value of the second binarization circuit and two functions of the second two-dimensional information calculation means according to a result of the dimensional information calculation means, and a background detection signal composed of a white detection signal as an input , N and m are arbitrary positive integers, the background information output in n × n pixel units is determined as one block unit, and the number of different white blocks is determined as a threshold for each m × m block plane. A fourth two-dimensional information calculating means having two functions for performing an operation, a second switching means for selecting and outputting one output of the two functions of the fourth two-dimensional information calculating means, Fifth two-dimensional information operation having a plurality of operation functions capable of performing S (S is a positive integer) different two-dimensional information operations in two ways on an m × m block plane by the output of the switching means Means and the fifth two-dimensional information calculating means. The character / photo area determination results obtained from the area determination result as the final determination result after S times and the two threshold value determination results of the fourth two-dimensional information calculation means are sequentially stored, and the current m × m block plane is stored. The background wider than the m × m block plane is obtained by using the character / photo area determination result obtained in the above and the character / photo area determination result obtained and stored from the area around the current m × m block plane. And a global area determining means for globally determining a character / photo area, and a second switching means and a fifth second means based on the determination result of the global area determining means.
An area determining circuit for switching a plurality of arithmetic functions of the dimension information calculating means and outputting an area determination result to the outside as an area switching signal; and a character binarizing signal or an identification pattern signal based on the area switching signal. An area switching circuit for selecting and outputting a character, and a pattern detection / separation circuit for discriminating a character binarization signal and an identification pattern signal coming from the area switching circuit and individually outputting them as a character binarization signal and a pattern detection signal. , And a black background is replaced with a white background by an identification pattern signal.

The optical character recognition device of the present invention includes a character recognition unit, an image input device, and a microprocessor for controlling the whole, which are interconnected via a system bus. Sensor that reads the image signal one-dimensionally, an AD converter that digitizes the image signal read by the image sensor, and normalization by performing logarithmic conversion processing, shading processing, and the like on the digitized form image signal A normalization circuit for obtaining a normalized signal; a binarization circuit for binarizing the normalized signal and outputting it as a character binary signal; a pattern generator for generating a specific identification pattern signal; A first binarization circuit for detecting dark black with a higher threshold (darker density) than a character binarization circuit for generating a character binarization signal; A second binarization circuit for background detection capable of switching between two thresholds, a lower threshold (lighter density) and a higher threshold, than the binarization circuit, and an output of the first binarization circuit is n × n It has a first two-dimensional information calculation means for performing black density determination in units of pixels, and two functions for determining the output of the second binarization circuit in white density determination in units of n × n pixels. And a second two-dimensional information calculating means for detecting the first two-dimensional information as a white detection signal.
A background detection circuit for performing an operation of switching between a threshold value of the second binarization circuit and two functions of the second two-dimensional information calculation means according to a result of the dimensional information calculation means, and a background detection signal composed of a white detection signal as an input , N and m are arbitrary positive integers, the background information output in n × n pixel units is determined as one block unit, and the number of different white blocks is determined as a threshold for each m × m block plane. A fourth two-dimensional information calculating means having two functions for performing an operation, a second switching means for selecting and outputting one output of the two functions of the fourth two-dimensional information calculating means, Fifth two-dimensional information operation having a plurality of operation functions capable of performing S (S is a positive integer) different two-dimensional information operations in two ways on an m × m block plane by the output of the switching means Means and the fifth two-dimensional information calculating means. The character / photo area determination results obtained from the area determination result as the final determination result after S times and the two threshold value determination results of the fourth two-dimensional information calculation means are sequentially stored, and the current m × m block plane is stored. The background wider than the m × m block plane is obtained by using the character / photo area determination result obtained in the above and the character / photo area determination result obtained and stored from the area around the current m × m block plane. And a global area determining means for globally determining a character / photo area, and a second switching means and a fifth second means based on the determination result of the global area determining means.
An area determining circuit for switching a plurality of arithmetic functions of the dimension information calculating means and outputting an area determination result to the outside as an area switching signal; and a character binarizing signal or an identification pattern signal based on the area switching signal. An area switching circuit for selecting and outputting a character, and a pattern detection / separation circuit for discriminating a character binarization signal and an identification pattern signal coming from the area switching circuit and individually outputting them as a character binarization signal and a pattern detection signal. The microprocessor, based on a pattern detection signal output from the image input device when detecting a non-dropout color character frame printed in a thick form on the form, includes a target character written in the character frame. The character recognition unit instructs the character recognition unit to cut out the character by the character recognition unit control signal, and recognizes the character. The unit takes in the character binarization signal in response to the input of the character recognition unit control signal, detects the cutout of the character, and selectively detects and removes the character frame which is not a dropout color printed thickly. It is.

[Action]

According to the character / photo region determination device of the present invention described above,
The background detecting means includes a first binarization circuit that detects dark black with a higher threshold (darker density) than a character binarization circuit that generates a character binarization signal, and a threshold (lower than a character binarization circuit). A second binarization circuit for background detection capable of switching between two thresholds, that is, a light density and a high threshold, and the output of the first binarization circuit is used to determine the black density in units of n × n pixels. A first two-dimensional information calculating means for performing the first two-dimensional information calculating means, and a second two-dimensional information calculating means having two functions for performing the white density determination in units of n × n pixels for the output of the second binarizing circuit. By switching the threshold of the second binarization circuit and the two functions of the second two-dimensional information calculating means according to the result of the first two-dimensional information calculating means, for example, in an area without characters, The level at which the average of the background white level can be detected as the threshold value for the background white level detection , The background white level detection threshold is set to a value equal to or greater than the peak value of the background white level in a region with a character, and for a background detection output in a region without a character, for example, 2 × 2 When there is no background as in a two-dimensional information calculation process using white pixels as a determination unit, a process of determining that there is no background is performed. In a region with a character, one white pixel is used as a determination unit of the presence or absence of the background. Even if the white level of the background is close to the threshold for character binarization, or if the white level of the background is higher than the threshold for character binarization such as the internal background of dense characters, the background is detected for characters and line drawings. It is possible to do.

According to the character / photo region determination device of the present invention described above, for example, the character binarized signal is input to the area switching means in the path of the i × i pixel two-dimensional information. A third two-dimensional information calculating means for determining the positions of the inner, outer and peripheral portions of the black image by performing the calculation and outputting the three-valued information; and a third means for switching the pseudo halftone signal and the character binary signal in pixel units. A first signal switching unit and a first to third two-dimensional information calculation unit, which determine whether a character region or a photograph region is a pixel unit, and a first signal switching unit according to the pixel unit region determination result. A character binarization signal processing means comprising pixel level area determination means for controlling the character / photograph area is provided, and the determination of the character / photograph area is performed twice for each pixel and for each block. Alternative or reinforcing means of judgment It is possible to reduce erroneous determination of the pseudo halftone signaling is black around the photo edges and character inside of a problem when performing edge enhancement processing to.

Further, according to the character / photo region determination device of the present invention described above, the region determination unit performs background information output in units of n × n pixels as one block unit, for each m × m block plane, A fourth two-dimensional information calculation unit having two functions for performing a determination operation using the number of different white blocks as a threshold, and a fourth two-dimensional information calculation unit for selectively outputting one of the two functions of the fourth two-dimensional information calculation unit A second switching means and a plurality of calculation functions capable of performing S different two-dimensional information calculations in two ways on an m × m block plane by using the output of the second switching means. A character information based on the dimension information calculation means, the area determination result which is the final determination result after S times of the fifth two-dimensional information calculation means, and the two threshold value determination results of the fourth two-dimensional information calculation means. A global area determining means for determining a photograph area; Configured, 2 of the second switching means and the fifth by the determination result of the global area determination means
By switching between the plurality of calculation functions of the dimension information calculation means, it is possible to accurately determine the character / photo region.

Further, the fifth two-dimensional information calculation means divides the S-number of different two-dimensional information calculations into m-pixel × m-line units on an m × m block plane by one matrix calculation circuit. By doing so, a two-dimensional information calculation processing method, which is one of the character / photo region block determination processing methods, can be realized with a relatively simple circuit configuration.

Further, according to the image input device of the present invention described above, the black area due to the background plate determined to be a photographic area is masked by the white level signal in the character binarization signal processing means. The black background by the captured background plate can be easily replaced with a white background, and signal processing after image input can be simplified.

According to the optical character recognition device of the present invention described above,
For example, a black character frame printed thickly on a form can be determined as a photographic area and selectively removed, so that, for example, to remove a character frame that hinders reading of a target character in the character frame. This eliminates the need for complicated preparations such as printing the character frame with the same dropout color as the illumination light at the time of reading, and can simplify the specifications of the form to be input.

Embodiment 1 FIG. 1 is a block diagram showing an example of a circuit configuration of a character / photograph area determination device according to an embodiment of the present invention.

The character / photo region determination device of the present embodiment is roughly divided into a background detection circuit 1 (background detection means), a character binarization signal processing circuit 2 (character binarization signal processing means), and an area determination circuit 3
(Area determining means) and area switching circuit 4 (area switching means)
It is composed of

Then, the character / photo region determination device inputs the character binarization signal 12 and the pseudo halftone signal 13 which are input from outside to the character binarization signal processing circuit 2 and the region switching circuit 4, and the input to the background detection circuit 1. The area switching circuit 4 outputs an adaptive binary signal 17 based on the video normalized signal 11 and the slice signal 10 to be processed.

In this case, the character binarization signal 12 is a signal obtained by binarizing the input image at a density in a unit of one pixel, and the pseudo halftone signal 13 is a binary gradation of the input image by dither processing. Signal.

The slice signal 10 is, for example, a signal from which an operator specifies the density of an image from the outside, and the video normalization signal 11
Is the video signal obtained by a given image input device.
This signal is a signal that has been subjected to conversion, logarithmically converted, and subjected to shading correction.

Further, the adaptive binarized signal 17 output from the character / photo region determination device selects and outputs one of the character binarized signal 12 and the pseudo halftone signal 13 in accordance with the characteristics of the input image. Signal.

In the case of the present embodiment, the background detection circuit 1 includes a plurality of binarization circuits 51 (first binarization circuit) and a binarization circuit 52 (first binarization circuit) to which the slice signal 10 and the video normalization signal 11 are input. 2) and a plurality of matrix operation circuits 61 (first two-dimensional information operation means) for outputting the black detection signal 14 and the white detection signal 15 to the character binarization signal processing circuit 2 and the area determination circuit 3. ) And the matrix operation circuit 62 (second 2
Dimension information calculation means).

Then, the binarization circuit 51 detects dark black from the input slice signal 10 and the video normalization signal 11 and outputs it to the matrix operation circuit 61.
An operation of performing black density determination and outputting the black detection signal 14 is performed by the function indicated by the arithmetic circuit number 61 in FIG.

Further, the binarization circuit 52 detects white of the background from the slice signal 10, the video normalization signal 11, and the black detection signal 14, and outputs the white to the matrix operation circuit 62.
Of the two functions shown in the arithmetic circuit number 62 in FIG. 13, the white density determination is performed by the function instructing the black detection signal 14,
An operation of outputting the white detection signal 15 is performed.

On the other hand, the character binarization signal processing circuit 2 outputs the character binarization signal.
A matrix operation circuit 63 to which 12 is input, a multi-level signal output from the matrix operation circuit 63 (third two-dimensional information operation means), and the black detection signal 14 and the white detection signal 15 are input. A selector 81 that performs predetermined processing on the character binary signal 12 and the pseudo halftone signal 13 arriving from the outside in accordance with the selector control circuit 71 and a control signal from the selector control circuit 71, and outputs the processed signal to the area switching circuit 4 ( First
Signal switching means).

The matrix operation circuit 63 is the operation circuit number 63 in FIG.
In accordance with the respective arithmetic expressions represented by, the pixel positions of the inside and outside of the black image are determined from the input character binary signal 12 and output to the selector control circuit 71, and the selector control circuit 71 An operation of outputting a determination signal to the selector 81 is performed by determining a character / photo region in pixel units from the determination result of the circumference, the black detection signal, and the white detection signal.

Further, the selector 81 receives the character binarization signal 12 and the pseudo halftone signal 13, and outputs one of the signals in units of pixels based on the determination of the character / photograph area in units of pixels.

FIG. 7 shows an example of the configuration of the selector control circuit 71. As shown in the figure, the selector control circuit 71 of the present embodiment is provided with an AND gate 71a which receives an inner judgment signal of the matrix operation circuit 63 and an inverted signal of the white detection signal 15 of the matrix operation circuit 62, and An OR gate 71b that receives an AND gate 71a and an outside determination signal of the matrix operation circuit 63 as inputs, and a peripheral detection signal of the matrix operation circuit 63 as a trigger input, and the white detection signal 15 of the matrix operation circuit 62
A hexadecimal counter 71c having a reset input as an input, an AND gate 71d having an input from the hexadecimal counter 71c and the black detection signal 14 of the matrix operation circuit 61 as an input, an output from the AND gate 71d, and a matrix operation An AND gate 71e that receives the inverted signal of the outside detection signal of the circuit 63 and the white detection signal 15 of the matrix operation circuit 62 as inputs,
A final stage OR gate 71f which receives the output of 1e and the OR gate 71b as inputs and outputs "1" to the selector 81 when a character binary signal is selected and "0" when a pseudo halftone signal is selected. Have been.

Then, the selection of the character binarization signal of the peripheral black pixel is suppressed, and the number of times of determination of the character binarization signal selection in an area where white background is not found is limited to two by the hexadecimal counter 71c. This is for removing the inner contour of the photograph.

The area determination circuit 3 includes a matrix calculation circuit 64 (fourth two-dimensional information calculation means), a selector 82 (second signal switching means), and a selector control circuit 72 (global area determination means).
And a matrix operation circuit 65 (fifth two-dimensional information operation means) composed of a plurality of stages of matrix operation circuits 651, 652, 653. The matrix operation circuit 64 receives the white detection signal 15 as an input, Further, density determination (hereinafter referred to as background density determination) is performed for each block. This background density determination is based on the arithmetic circuit number in FIG.
Performed on two different functions, as shown at 64,
A selector control circuit for determining the result of each of the two functions
Output to 72 and selector 82.

The selector 82 switches the output from the matrix operation circuit 64 in the preceding stage according to the output of the selector control circuit 72. Further, the processing results of the two functions of the background density determination are
The matrix operation circuit 651 selected at 82
Is input to The matrix operation circuit 651 performs a binary judgment (character area judgment) process which acts to increase the background density judgment by the operation indicated by the operation circuit number 651 in FIG. Circuit 65
Output to 2.

Matrix operation circuit 652 and matrix operation circuit 6
53 performs a collection process on the binary determination result.

That is, the matrix operation circuit 652 and the matrix operation circuit 653 have two different functions corresponding to the operation circuit numbers 652 and 653 in FIG. 14, respectively, and the function operated by the output of the selector control circuit 72 is switched.

The matrix operation circuit 653 at the last stage outputs the region switching signal 16 to the region switching circuit 4 and the selector control circuit 72 as a final determination result.

The selector control circuit 72 makes a global character / photo area determination based on the two function outputs of the matrix operation circuit 64 and the area switching signal 16, and outputs a determination result.

FIG. 8 shows an example of the configuration of the selector control circuit 72. As shown in FIG.
Reference numeral 72 denotes an AND gate 72a that receives the area switching signal 16 that is the output of the last stage of the matrix operation circuit 653 and the output of the function 2 of the matrix operation circuit 64, the output of the AND gate 72a, and the output of the AND gate 72a. An AND gate 72d which receives an output from the plurality of registers 72b and 72c as an input, an output of the AND gate 72d as a set input, and an output of the function 1 of the matrix operation circuit 64 as a reset input to the selector 82 as "1" or An SR flip-flop 72e that outputs a logical control signal of “0” is provided. The SR flip-flop 72e outputs a logical control signal of “0”. Based on this, a global character / photo region is determined as described below.

The area switching circuit 4 includes a line memory 91 in which the incoming pseudo halftone signal 13 is stored, and a processed signal arriving via the selector 81 in the character binarized signal processing circuit 2 in the preceding stage.
A line memory 92 for storing 12a and an output from these line memories 91 and 92 are selected to generate an adaptive binary signal 17
And a selector 83 for outputting to the outside.

That is, the line memory 91 and the line memory 92 function to delay the pseudo halftone signal 13 and the character binarization signal 12 by an amount corresponding to the determination signal by the processing in the background detection circuit 1 and the area determination circuit 3.

The pseudo halftone signal 13 is input to the line memory 91, but the line memory 92 is determined not to be the character binary signal 12 itself but to be a photographic area in pixel units based on pixel area determination. And the processed signal 12a switched to the pseudo halftone signal 13 by the selector 81.

Next, an embodiment of a general matrix operation circuit 6000 constituting the matrix operation circuits 61, 62, 63, 64 and 65 is shown in FIG. In addition, in order to simplify the description, 5 ×
5 shows a case where a two-dimensional information calculation with a matrix size of 5 is performed.

As shown in FIG.
Is 25 registers 60 corresponding to the number of elements in the matrix
01, line memory 6002 for 4 lines, and 1st and 14th
Operation ROM 6003 that performs two-dimensional information operation as shown in the figure
Consists of The operation ROM 6003 can switch the operation function by the function switching signal, and can simultaneously perform three two-dimensional information operation processes.

That is, in the case of the matrix operation circuit 63 of the area determination circuit 3 as an example, a plurality of matrix operations in block units may be performed once for each block unit, and the block size is n pixels × n lines. In this case, one operation may be performed once every n lines, and if a matrix operation is time-division-processed in line units, n two-dimensional information operations can be performed by one matrix operation circuit (however, a plurality of operation outputs are provided). .

FIG. 10 shows a multi-stage matrix operation circuit 65 constituting the matrix operation circuit 65 in the area determination circuit 3.
This shows an example of a configuration in which one matrix operation circuit 6000 implements three matrix operation processes in 1,652,653.

That is, the matrix operation circuit 65 in the present embodiment
Is a matrix operation circuit 6000 as shown in FIG. 9 for simultaneously executing three operation processes, a timing control circuit 65a for selecting an output from the matrix operation circuit 6000, a plurality of logic gates 65b and 65c, and an operation result. Is stored for one line, and a line memory 65e for holding and outputting the final calculation result output once every n lines for n lines.

Thereby, the matrix operation processing of the matrix operation circuit 65 as shown in FIG. 14 can be realized without increasing the circuit scale.

 Hereinafter, the operation of the present embodiment will be described.

Normally, the threshold for binarizing characters is PCS (Print Cnotrast Si
gnal) value is between 0.10 and 0.50, and an operator from outside selects this value by specifying the density on the operation panel. In most cases, when the density of a character is viewed as a peak value, the PCS value is 0.5 or more.

On the other hand, when the white level of the background is considered by the PCS value, it depends on the presence or absence of the automatic adjustment of the density. Looking at the average white level of the background, PCS 0.0
The value is 5 or lower. As a characteristic of the background detection circuit used for area determination, if one of the conditions for character area determination is that white background is found in the periphery, the background is not found as much as possible in areas where there is no character. Making the background as easy to find as possible in the region gives good results in the region determination. In this case, the region with the character means a region where the possibility of the character is large, and in this embodiment, a case where dark black is found, and this dark black is a PCS that can be distinguished from the background PCS peak value Think value, PCS
It refers to a concentration of 0.5 or more (for example, PCS 0.6).

Therefore, in the present embodiment, the following is performed to obtain a good area determination result.

That is, in the binarization circuits 51 and 52 of the background detection circuit 1, the background white level detection threshold value in an area without a character is reduced to a level at which the average of the white levels of the background can be detected, and the background level in the area with a character is detected. The white level detection threshold value is set to, for example, a PCS value of 0.5 as a value equal to or higher than the peak value of the white level of the background. FIG. 6 is a graph showing the relationship between the above-mentioned detection thresholds. The matrix operation circuits 61 and 62 perform processing for determining that there is no background when there is no background in a concentrated manner with respect to the output of background detection in an area without characters, for example, 2
A two-dimensional information calculation process is performed using a × 2 white pixel as a determination unit, and an operation is performed using one white pixel as a determination unit in an area with a character.

Thereby, even when the white level of the background is close to the threshold value of character binarization such as a color-printed original, or when the white level of the background becomes higher than the threshold value of character binarization such as the internal background of dense characters such as kanji, For characters and line drawings, the background can be reliably detected.

On the other hand, when the character / photo area is switched in units of pixel blocks of the processed image, the area is set to 2 in the middle of the thick line drawing.
May be divided. Therefore, in the present embodiment, for such a thick line drawing, even if it is a character area, the area inside the thick line is converted into a photographic area to resolve the area division into two.

That is, as means for internalizing the thick line, the matrix operation circuit 63 of the character / binarization signal processing circuit 2 uses
By the matrix operation of 5 × 5 pixels shown in FIG. 13,
The determination is made by looking at the black distribution of the character binary signal.
That is, as shown in FIG.
63 designates a region A as a region where no black exists in a matrix of 5 × 5 pixels for a thick line image, and a region where a central pixel of 5 × 5 pixels is white and peripheral pixels have black. B is determined, a region C is determined as a region in which the central pixel of 5 × 5 pixels is black and peripheral pixels are white, and 1 × 5 × 5 pixels
The areas D and E are determined as areas having no white. Next, an area D where the white of the background is found in the 8 × 8 pixels and an area where the white of the background is not found are obtained by the white detection signal 15 of the matrix operation circuit 62 for detecting the white of the 8 × 8 pixel matrix size background. And the area E is determined and separated.

Then, the region A and the region E are set as the photograph region, and the region
Set B, C, and D to the character area.

If the thickness of the line drawing is 8 pixels or less (0.5 mm when the line density is 16 lines / mm), the entire line drawing is determined to be a character area, but if the line drawing is 8 pixels or more, the thickness is 4 pixels Is determined as a character area, and the area inside the outline is determined as a photograph area. For a line drawing that is thicker than a certain extent (2 mm square in the present embodiment), the whole is determined to be a photographic region by character / photo region determination in block units. In this way, thickness 0.5mm
A normal character of the degree is a character region, a character or line drawing having a thickness of 0.5 mm to 2 mm is a photographic region having an outline, and a photographic image having a larger thickness is a photographic region having no outline.

In this way, as an alternative or reinforcing means for character / photo area determination, an edge around the black area of the photo area, which becomes a problem when performing contour emphasis processing, or a pseudo halftone signal inside the character is generated. This makes it possible to reliably prevent erroneous determinations such as mistakes.

The block determination process in the area determination circuit 3 is as follows.

First, as shown in FIG. 14, the matrix operation circuit 64 receives the 8 ×
The background density is determined by the function 1 and the function 2 on a 4 × 4 block plane in which the two-dimensional information calculation output of the background white detection of the block of 8 pixels is one block. This determination result is output in units of one block (every eight pixels and eight lines). The output described below is also output in units of one block.

Next, the matrix operation circuit 651 performs a binary judgment on the result of the background density judgment of the matrix operation circuit 64 in the preceding stage on a 4 × 4 block plane. Next, with respect to the result of the binary determination, correction processing (smoothing processing) is performed twice on a 3 × 3 block plane in matrix operation circuits 652 and 653 at the subsequent stage, and the final result is obtained as an area switching signal 16 in block units. The signal is output to the selector 83 of the area switching circuit 4 and is input to the selector control circuit 72 at the same time.

Here, the matrix operation circuit 64 and the matrix operation circuits 652 and 653 have two functions 1 and 2 respectively for the background density determination and the correction processing as described above. If the result of the target area determination is a character area, function 1 is selected, and if it is a photograph area, function 2 is selected.

The global area determination in the selector control circuit 72 is performed based on two function outputs of the background density determination and the final determination result.
Function 1 for determining background density in matrix operation circuit 64
When is not true, the photograph area is globally considered. (In this embodiment, the SR flip-flop 72 shown in FIG. 8 is used.
e is reset, and a logic signal of “0” level is output to the selector 82 and the matrix operation circuits 652 and 653. On the other hand, when the function 2 of the background density determination in the matrix operation circuit 64 is true and the final determination result is true (character area) for two blocks (32 pixels), the character area is globally defined as a trigger ( In this embodiment, the SR flip-flop 72e shown in FIG. 8 is set, and a logic signal of "1" level is output to the selector 82 and the matrix operation circuits 652 and 653).

That is, in the present embodiment, the global area determination by the selector control circuit 72 is immediately determined to be a global photographic area when black of 2 mm or more is detected. There is a property of returning to the global character area, but if a photographic area (block determination result) is around, even if white is 2 mm or more, the white area is still globally determined to be a photographic area, Erroneous determination is less likely to occur even when a relatively large white area exists in the photograph area.

Next, the background density determination in the matrix operation circuit 64 will be described with reference to output examples of two functions.

In the case of black smaller than 2 mm square as shown in FIG. 3 and larger than 1 mm square as shown in FIG. 4, the attention determination block at the center (B in FIG. 0. As shown in FIG. 4, in the case of black of 1 mm or less, the attention judgment block at the center is 1 for both functions 1 and 2. Black is 2
mm or more, the attention judgment block at the center
Both functions 1 and 2 are 0. As described above, the function 1 and the function 2 make the same judgment when the black is 2 mm or more and 1 mm or less, and make different judgments when the black has a size of 1 to 2 mm. In order to make it easier to determine an area having a character as a character area, a function 1 for background density determination is selected based on a global determination result. In order not to use the white area of the photo area without text as the text area,
The function 2 of the background density determination is selected based on the global determination result.

Next, the process of the binary determination in the matrix operation circuit 651 will be described. The determination by the function 1 of the background density determination is a kind of black thinning processing, which can be said to be a processing of making a character area (meaning that there is a background) up to 1 mm inward of black. do.
That is, as shown in FIG. 5, a process of expanding the character area is performed by taking the condition that there are white backgrounds at the four corners.

Next, the collection processing by the matrix operation circuits 652 and 653 will be described. As shown in FIG. 14, the majority process is performed in the 3 × 3 block plane. Function 1 and Function 2 change the majority decision threshold. Function 1 converts an isolated one-block size photo area into a character area, and Function 2 converts an isolated 3 × 3 block-size character area into a photo area. I do. And the selector control circuit 72
When a character area is selected according to the global determination result, the function 1 is selected, and the erroneous determination of the photograph area is corrected.

When a photograph area is selected based on the global determination result of the selector control circuit 72, the function 2 is selected, and the character area erroneous determination is corrected.

In this manner, the character / photo region can be accurately determined by the block determination without being affected by noise or the like.

Further, in the case of the present embodiment, the matrix operation circuits 651, 652, and 653 described above include one matrix operation circuit 6000 and a timing control circuit 65a as shown in FIG.
And a plurality of logic gates 65b and 65c and a plurality of line memories 6
5d and 65e are realized by a set of simple circuit configurations, and a plurality of matrix operation processes as described above can be performed without unnecessarily increasing the circuit scale.

Second Embodiment Next, a description will be given of an embodiment in which the character / photograph area determination device according to one embodiment of the present invention is incorporated in an image input device and used.

That is, the image input device 100 shown in FIG. 11 includes an image sensor 103 that reads image data one-dimensionally from a target form 101, and an image sensor 103 that is disposed behind the form 101 and has an end portion in the width direction of the form 101. And a black background plate 102 capable of detecting an image.By sending the form 101 at a predetermined speed in a direction perpendicular to the reading width direction of the image sensor 103, the entire area of the form 101 is input as a two-dimensional image. Structure.

The image sensor 103 includes an AD converter 104 for digitizing a read image signal, and a well-known logarithmic conversion process and shading process performed on the digitized image signal of the form 101 to generate a normalized signal. 119
And a normalization circuit 105 for obtaining

Further, the normalized signal 119 is supplied to a predetermined binarizing circuit 106
Is inputted to the line memory 92 of the area switching circuit 4 having the structure as shown in the first embodiment, and a part of the signal is inputted to the background detection circuit 1.

In this case, the pattern is generated in the area switching circuit 4 by the pattern generator 107 instead of the pseudo halftone signal. A predetermined identification pattern signal 120 different from the character binary signal 123 is input to the line memory 91.

Although not specifically shown, the background detection circuit 1 receives a slice level signal (not shown) set externally by an operator or the like.

In response to the background detection signal 121 output from the background detection circuit 1, the area determination circuit 3 sends an area switching signal 122 to the area switching circuit 4 to switch between the character binarization signal 123 and the identification pattern signal 120. It is configured to perform an operation.

In the subsequent stage of the area switching circuit 4, a pattern detection / separation circuit 11
1 is connected to discriminate the character binary signal 123 and the identification pattern signal 120 coming from the area switching circuit 4,
An operation of outputting as a character binarization signal 112 and a pattern detection signal 113 is performed.

By the way, as described in the first embodiment, the character binarization signal 112 is different from the character binarization signal 123 coming from the binarization circuit 106 in that the region switching signal 122 is determined to be a photographic region. , Are masked by the white level signal. That is, the black portion of the captured image due to the black background plate 102 has no white background detected and is determined to be a photographic region. Therefore, according to this embodiment, in the conventional apparatus, the black background (the black background plate portion) can be replaced with the white background by the internal processing while the black background plate 102 is taken as black as it is,
Signal processing in a higher-level processing device after image input is simplified.

In addition, since a sufficiently large block of black pixels on the form 101 is determined to be a photograph area, a character frame that is not a dropout color and is printed thickly on the form 101 is determined to be a photograph area, and is removed from the captured image signal (drop). Out) can be.

That is, in a specific reading area of the form 101,
In order to specify the data writing position, it is visible to the operator who actually performs the writing operation, and in order to drop out at the time of reading, the writing frame etc. is conventionally printed with the same color as the illumination of the reading optical system with the dropout color. This eliminates the necessity of printing, and it suffices to simply print a thick writing frame in normal black, which is the same as the actual writing character, and the specifications of the form 101 are simplified.

Third Embodiment FIG. 12 shows an example in which the pattern detection signal 113 output from the image input device 100 in the second embodiment is applied to an optical character recognition device.

As shown in the figure, the optical character recognition device 200 of the third embodiment includes, on a predetermined system bus 215, the above-described image input device 100, a character recognition unit 216, and a microprocessor 217 for performing overall control. Are connected.

A character frame that is not a dropout color is printed thick on the form 101 input to the image input device 100, and the image input device 100 has detected the character binarization signal 112 obtained by dropping out the character frame and the character frame. The pattern detection signal 113 is output to the system bus 215.

Then, the microprocessor 217 knows the cutout position of the target character written in the character frame from the pattern detection signal 113 output upon detection of the character frame, and cuts out the character by the character recognition unit control signal 218. To the character recognition unit 216. In response to this, the character recognition unit 216 takes in the character binarization signal 112 and performs character cutout detection.

In this case as well, in a specific reading area of the form 101, the operator of the actual writing operation is visible to specify the writing position of the data, and for the purpose of dropping out at the time of reading, conventionally, the illumination of the reading optical system is used. It is not necessary to print the writing frame etc. with the same color dropout color as it is, just print the thick black writing frame with the same normal black as the actual writing character, and the specifications of the form 101 will be simplified .

Although the invention made by the inventor has been specifically described based on the embodiment, the present invention is not limited to the embodiment, and it is needless to say that various modifications can be made without departing from the gist of the invention. Nor.

For example, circuit configurations such as a background detection circuit, an area determination circuit, an area switching circuit, a character binarization signal processing circuit, and algorithms for various operations are not limited to those exemplified in the above-described embodiment.

〔The invention's effect〕

The effects obtained by typical aspects of the invention disclosed in the present application will be briefly described as follows.

That is, according to the character / photo region determination device of the present invention, the background is detected by two-dimensional information calculation and output as binary background information, and the character / photo region is calculated by two-dimensional information calculation from the background information. Based on an area determining means for determining a photograph area and an area determination result output from the area determining means, a character binarized signal, which is externally received and is binarized with a certain threshold, and a threshold for each pixel are set. And a region switching means for selecting and outputting any of the pseudo halftone signals binarized by changing the signal.
A first binarization circuit that detects dark black with a higher threshold (darker density) than the binarization circuit, and two thresholds of a lower threshold (lighter density) and a higher threshold than the character binarization circuit can be switched. A second binarizing circuit for background detection;
A first two-dimensional information calculation unit that performs black density determination on the output of the binarization circuit in units of n × n pixels, and performs a white density determination on the output of the second binarization circuit in units of n × n pixels A second two-dimensional information calculation means having two functions,
By performing the operation of switching between the threshold of the second binarization circuit and the two functions of the second two-dimensional information calculating means according to the result of the two-dimensional information calculating means, for example, in a region without characters, a white background Reduce the level detection threshold to a level at which the average of the background white level can be detected, and set the background white level detection threshold to a value greater than or equal to the peak value of the background white level in areas with characters, In the case of the background detection output in step (2), for example, a process of determining that there is no background when there is no concentrated background, such as a two-dimensional information calculation process using 2 × 2 white pixels as a determination unit, is performed. In a certain area, one white pixel is used as a unit for determining the presence or absence of a background, so that the white level of the background is close to the threshold value of the character binarization, or the character binarization such as the internal background of a dense character is performed. Even if it ’s higher than the threshold It is possible to detect the background with respect to a character or line drawing.

According to the character / photo region determination device of the present invention described above, for example, the character binarized signal is input to the area switching means in the path of the i × i pixel two-dimensional information. A third two-dimensional information calculating means for determining the positions of the inner, outer and peripheral portions of the black image by performing the calculation and outputting the three-valued information; and a third means for switching the pseudo halftone signal and the character binary signal in pixel units. A first signal switching unit and a first to third two-dimensional information calculation unit, which determine whether a character region or a photograph region is a pixel unit, and a first signal switching unit according to the pixel unit region determination result. A character binarization signal processing means comprising pixel level area determination means for controlling the character / photograph area is provided, and the determination of the character / photograph area is performed twice for each pixel and for each block. Alternative or reinforcing means of judgment It is possible to reduce erroneous determination of the pseudo half signaling is black around the photo edges and character inside of a problem when performing edge enhancement processing to.

Further, according to the character / photo region determination device of the present invention described above, the region determination unit performs background information output in units of n × n pixels as one block unit, for each m × m block plane, A fourth two-dimensional information calculation unit having two functions for performing a determination operation using the number of different white blocks as a threshold, and a fourth two-dimensional information calculation unit for selectively outputting one of the two functions of the fourth two-dimensional information calculation unit A second switching means and a plurality of calculation functions capable of performing S different two-dimensional information calculations in two ways on an m × m block plane by using the output of the second switching means. A character information based on the dimension information calculation means, the area determination result which is the final determination result after S times of the fifth two-dimensional information calculation means, and the two threshold value determination results of the fourth two-dimensional information calculation means. A global area determining means for determining a photograph area; Configured, 2 of the second switching means and the fifth by the determination result of the global area determination means
By switching between the plurality of calculation functions of the dimension information calculation means, it is possible to accurately determine the character / photo region.

Further, the fifth two-dimensional information calculation means divides the S-number of different two-dimensional information calculations into m-pixel × m-line units on an m × m block plane by one matrix calculation circuit. By doing so, a two-dimensional information calculation processing method, which is one of the character / photo region block determination processing methods, can be realized with a relatively simple circuit configuration.

Further, according to the image input device of the present invention, in an image input device that uses a black or gray background plate to detect a form, an image sensor that reads one-dimensional image data from a target form, An AD converter that digitizes the image signal read by the image sensor, a normalization circuit that performs logarithmic conversion processing and shading processing on the digitized image signal of the form to obtain a normalized signal, Binarization circuit that binarizes the digitized signal and outputs it as a character binarized signal, a pattern generator that generates a specific identification pattern signal, and a character that receives a normalized signal and generates a character binarized signal A first binarization circuit that detects dark black with a higher threshold (darker density) than the binarization circuit, and a lower threshold (lighter density) and a higher threshold than the character binarization circuit A second binarization circuit for background detection capable of switching between two thresholds, and a first two-dimensional information calculation for determining the output of the first binarization circuit as a black density in units of n × n pixels Means, and two functions for performing white density determination on the output of the second binarization circuit in units of n × n pixels, and detecting the background white and outputting the same as a white detection signal. The first 2
A background detection circuit for performing an operation of switching between a threshold value of the second binarization circuit and two functions of the second two-dimensional information calculation means according to a result of the dimensional information calculation means, and a background detection signal composed of a white detection signal as an input , N and m are arbitrary positive integers, the background information output in n × n pixel units is determined as one block unit, and the number of different white blocks is determined as a threshold for each m × m block plane. A fourth two-dimensional information calculating means having two functions for performing an operation, a second switching means for selecting and outputting one output of the two functions of the fourth two-dimensional information calculating means, Fifth two-dimensional information operation having a plurality of operation functions capable of performing S (S is a positive integer) different two-dimensional information operations in two ways on an m × m block plane by the output of the switching means Means and the fifth two-dimensional information calculating means. The character / photo area determination results obtained from the area determination result as the final determination result after S times and the two threshold value determination results of the fourth two-dimensional information calculation means are sequentially stored, and the current m × m block plane is stored. The background wider than the m × m block plane is obtained by using the character / photo area determination result obtained in the above and the character / photo area determination result obtained and stored from the area around the current m × m block plane. And a global area determining means for globally determining a character / photo area, and a second switching means and a fifth second means based on the determination result of the global area determining means.
An area determining circuit for switching a plurality of arithmetic functions of the dimension information calculating means and outputting an area determination result to the outside as an area switching signal; and a character binarizing signal or an identification pattern signal based on the area switching signal. An area switching circuit for selecting and outputting a character, and a pattern detection / separation circuit for discriminating a character binarization signal and an identification pattern signal coming from the area switching circuit and individually outputting them as a character binarization signal and a pattern detection signal. Since a black background is replaced with a white background by the identification pattern signal, signal processing after image input can be simplified.

Further, the optical character recognition device according to the present invention includes a character recognition unit, an image input device, and a microprocessor for controlling the whole, which are interconnected via a system bus. An image sensor that reads data one-dimensionally, an AD converter that digitizes the image signal read by the image sensor, and a logarithmic conversion process and a shading process that are performed on the digitized form image signal A normalization circuit for obtaining a normalized signal; a binarization circuit for binarizing the normalized signal and outputting it as a character binary signal; a pattern generator for generating a specific identification pattern signal; A first binarization circuit that detects dark black with a higher threshold (darker density) than a character binarization circuit that generates a character binarization signal; A second binarization circuit for background detection that can switch between two thresholds, a lower threshold (lighter density) and a higher threshold, than the character binarization circuit, and the output of the first binarization circuit is nx It has a first two-dimensional information calculating means for performing black density determination in units of n pixels, and two functions for performing white density determination in units of n × n pixels by using the output of the second binarization circuit. A second two-dimensional information calculating means for detecting white and outputting it as a white detection signal;
A background detection circuit for performing an operation of switching between a threshold value of the second binarization circuit and two functions of the second two-dimensional information calculation means according to a result of the dimensional information calculation means, and a background detection signal composed of a white detection signal as an input , N and m are arbitrary positive integers, the background information output in n × n pixel units is determined as one block unit, and the number of different white blocks is determined as a threshold for each m × m block plane. A fourth two-dimensional information calculating means having two functions for performing an operation, a second switching means for selecting and outputting one output of the two functions of the fourth two-dimensional information calculating means, Fifth two-dimensional information operation having a plurality of operation functions capable of performing S (S is a positive integer) different two-dimensional information operations in two ways on an m × m block plane by the output of the switching means Means and the fifth two-dimensional information calculating means. The character / photo area determination results obtained from the area determination result as the final determination result after S times and the two threshold value determination results of the fourth two-dimensional information calculation means are sequentially stored, and the current m × m block plane is stored. The background wider than the m × m block plane is obtained by using the character / photo area determination result obtained in the above and the character / photo area determination result obtained and stored from the area around the current m × m block plane. And a global area determining means for globally determining a character / photo area, and a second switching means and a fifth second means based on the determination result of the global area determining means.
An area determining circuit for switching a plurality of arithmetic functions of the dimension information calculating means and outputting an area determination result to the outside as an area switching signal; and a character binarizing signal or an identification pattern signal based on the area switching signal. An area switching circuit for selecting and outputting a character, and a pattern detection / separation circuit for discriminating a character binarization signal and an identification pattern signal coming from the area switching circuit and individually outputting them as a character binarization signal and a pattern detection signal. The microprocessor, based on a pattern detection signal output from the pixel input device when detecting a character frame that is not a dropout color printed on a form thickly, based on a target character written in the character frame. The character recognition unit instructs the character recognition unit to cut out the character by the character recognition unit control signal, and recognizes the character. The unit takes in the character binarization signal in response to the input of the character recognition unit control signal, and detects and cuts out the character, thereby selectively detecting and removing a character frame that is not a thickly printed dropout color. For example, a black character frame printed thickly on a form can be determined as a photograph area and selectively removed, for example, for the purpose of removing a character frame obstructing reading of a target character in the character frame. ,
This eliminates the need for complicated preparations such as printing the character frame with the same dropout color as the illumination light used for reading, and can simplify the specifications of the form to be input.

[Brief description of the drawings]

FIG. 1 is a block diagram of a character / photo region determination apparatus according to one embodiment of the present invention, FIG. 2 is an explanatory diagram of pixel-based region determination, FIG. 3 is an explanatory diagram of background density determination processing, FIG. FIG. 5 is an explanatory diagram of the background density determination process, FIG. 5 is an explanatory diagram of the binary determination process, FIG. 6 is a graph showing an example of the setting of the detection threshold in the background detection circuit, and FIG. FIG. 8 is a block diagram illustrating an example of a circuit configuration of a selector control circuit that performs a global area determination. FIG. 8 is a block diagram illustrating an example of a circuit configuration of a selector control circuit that performs a global area determination. FIG. 10 is a block diagram showing a configuration of a matrix operation circuit in which a plurality of two-dimensional information operations are performed by one matrix operation circuit. FIG. 11 is a block diagram showing an example of the configuration of an image input device. The figure shows optical character recognition FIG. 13 is a block diagram showing an example of the configuration of the device, FIG. 13 is a diagram showing an example of the functions of the arithmetic expressions and functions of the two-dimensional information operation in the character / photograph area determination device according to one embodiment of the present invention, and FIG. FIG. 5 is a diagram illustrating an example of the functions of the operation formulas and functions of the two-dimensional information calculation in the character / photograph area determination device according to the embodiment of the present invention. 1... Background detection circuit (background detection means), 51... Binarization circuit (first binarization circuit), 52... Binarization circuit (second binarization circuit)
... Matrix operation circuit (first two-dimensional information operation means), 62... Matrix operation circuit (second
, 2... Character binarization signal processing circuit (character binarization signal processing means), 63... Matrix operation circuit (third 2D information calculation means), 71. Circuit, 71a …… AND gate, 71b …… OR gate, 71c…
... hexadecimal counter, 71d, 71e ... AND gate, 71f ... OR gate, 81 ... selector (first signal switching means), 3 ...
Area judgment circuit (area judgment means), 64... Matrix operation circuit (fourth two-dimensional information operation means), 651 to 653... Matrix operation circuit, 65... Matrix operation circuit (fifth two-dimensional information operation means) ), 65a: timing control circuit, 65b, 65c: logic gate, 65d, 65e: line memory, 72: selector control circuit, 82: selector (second signal switching means), 4: area switching Circuit (area switching means), 83 ... selector, 91,92 ... line memory, 10 ...
... Slice signal, 11 ... Video normalization signal, 12 ... Character binarization signal, 12a ... Processing signal, 13 ... Pseudo halftone signal, 14 ... Black detection signal, 15 ... White detection signal, 16 ………………………………………………………………………………………………………………………………………… 100 100 100 100, 101 100 100 100 100 100 100 100 103 103 103 103. ... Normalization circuit, 106
... Binarization circuit, 107 pattern generator, 111 pattern detection / separation circuit, 112 character binarization signal, 113 pattern detection signal, 119 normalized signal, 120 identification pattern Signal, 121... Background detection signal, 122... Area switching signal, 123... Character binary signal, 200... Optical character recognition device, 215... System bus, 216.
217: microprocessor, 218: character recognition unit control signal.

 ──────────────────────────────────────────────────の Continuing on the front page (72) Inventor Yasutomi Ejiri 2880 Kozu, Kozuhara, Odawara City, Kanagawa Prefecture Inside the Odawara Plant, Hitachi, Ltd. (56) References JP-A-58-3374 (JP, A) JP-A-62-145968 (JP, A)

Claims (5)

(57) [Claims] A background is detected by two-dimensional information calculation, and a background is detected.
Background detecting means for outputting as value background information; area determining means for determining a character / photograph area by two-dimensional information calculation from the background information; Area switching means for selecting and outputting one of a character binary signal binarized with a constant threshold and a pseudo halftone signal binarized by changing the threshold for each pixel. Wherein the background detection means comprises: a first binarization circuit for detecting dark black with a higher threshold (darker density) than a character binarization circuit for generating the character binarization signal; A second binarization circuit for background detection capable of switching between two thresholds, a lower threshold (lighter density) and a higher threshold, and an output of the first binarization circuit of n × n pixels. Perform black density judgment as a unit A first two-dimensional information computing means, the second
And a second two-dimensional information calculating means which has two functions for performing white density determination on the output of the binarizing circuit in units of n × n pixels, and detects a white background and outputs it as a white detection signal. Performing an operation of switching a threshold value of the second binarization circuit and the two functions of the second two-dimensional information calculating means according to a result of the first two-dimensional information calculating means; When n and m are arbitrary positive integers, the background information output in units of n × n pixels is determined as one block unit, and the number of different white blocks is determined as a threshold value for each m × m block plane. A fourth two-dimensional information calculating means having two functions for performing operations, a second switching means for selectively outputting one of the two functions of the fourth two-dimensional information calculating means, The output of the switching means is an m × m block plane Fifth two-dimensional information calculation means including a plurality of calculation functions capable of performing S (S is a positive integer) different two-dimensional information calculations in two ways, and a fifth two-dimensional information calculation means The character / photo region determination results obtained from the region determination result as the final determination result after S times and the two threshold value determination results of the fourth two-dimensional information calculation means are sequentially stored, and the current m × m Using the character / photo area determination result in the block plane of the above and the text / photo area determination result obtained from the area around the current m × m block plane and stored, the m × m A global area determining means for globally determining a character / photo area from a background wider than the block plane of the second switching means and the fifth two-dimensional information based on the determination result of the global area determining means. The plurality of functions of the arithmetic means Character-photograph region determining apparatus being characterized in that to perform the switching function. 2. The character 2 input to the area switching means.
In the path of the binarized signal, the character binarized signal is subjected to two-dimensional information operation of i × i pixels, and the case where all of the i × i pixels are white is outside the black image, and all of the i × i pixels are In the case of black, the center pixel of i × i pixels is white and the surrounding black is 1
And the center pixel of i × i pixels is black and there is at least one surrounding white pixel, it is determined to be the peripheral part of the black image, thereby determining the position of the inner, outer and peripheral parts of the black image. A third two-dimensional information calculating unit that outputs a ternary signal, a first signal switching unit that switches the pseudo halftone signal and the character binary signal on a pixel-by-pixel basis, the first, second, and third signals. Characters or pixel areas are determined on the basis of the result of the two-dimensional information calculation means in a pixel unit, and the character is composed of a pixel level area determination means for controlling the first signal switching means in accordance with the area determination result on a pixel basis. 2. The character / photograph area judging device according to claim 1, wherein a binarized signal processing means is provided. 3. The method according to claim 2, wherein the fifth two-dimensional information calculation means is m × m
By dividing the S different two-dimensional information operations into m pixel × m line units on the block plane of
2. The character / photograph area determination device according to claim 1, wherein the determination is performed by one matrix operation circuit. 4. An image input device that uses a black or gray background plate to detect a form of a form, comprising: an image sensor that reads image data one-dimensionally from a target form; An AD converter that digitizes the digitized image signal; a normalization circuit that performs a logarithmic conversion process, a shading process, and the like on the digitized image signal of the form to obtain a normalized signal; A binarization circuit that binarizes the signal and outputs it as a character binarization signal; a pattern generator that generates a specific identification pattern signal; and the normalization signal is input to generate the character binarization signal A first binarization circuit for detecting dark black with a higher threshold (darker density) than the character binarization circuit, and a lower threshold (lighter density) and a higher threshold than the character binarization circuit A second binarization circuit for background detection capable of switching between the two threshold values, and a first 2 which performs black density determination on the output of the first binarization circuit in units of n × n pixels. A second information processing means for calculating a white density of the output of the second binarization circuit in units of n × n pixels, and detecting a white background and outputting the same as a white detection signal; An operation of switching the threshold of the second binarization circuit and the two functions of the second two-dimensional information calculating means according to the result of the first two-dimensional information calculating means. And a background detection signal comprising the white detection signal, and when n and m are arbitrary positive integers, the background information output in n × n pixel units is converted into one block unit. As m
A fourth two-dimensional information calculating means having two functions for performing a determination operation using the number of different white blocks as a threshold value for each of the × m block planes, and a function of the four functions of the fourth two-dimensional information calculating means. A second switching means for selecting and outputting one of the outputs, and performing an S (S is a positive integer) number of different two-dimensional information calculations on the output of the second switching means in two ways on an m × m block plane. Fifth with a plurality of arithmetic functions capable of
A dimension information calculation unit, a character obtained from an area determination result as a final determination result after S times of the fifth two-dimensional information calculation unit, and two threshold value determination results of the fourth two-dimensional information calculation unit. The photograph area determination results are sequentially stored, and the current m
Using the character / photo area determination result on the × m block plane and the character / photo area determination result obtained from the area around the current m × m block plane and stored, A global area determining means for globally determining a character / photograph area from a background wider than a block plane of × m, and the second switching means and the fifth and second elements are determined by the determination result of the global area determining means. An area determination circuit that switches the plurality of arithmetic functions of the dimension information calculation means and outputs the area determination result to the outside as an area switch signal; and based on the area switch signal, the character binary signal and the An area switching circuit for selecting and outputting any one of the identification pattern signals; and a character binarization signal discriminating a character binary signal and an identification pattern signal coming from the area switching circuit. No. and pattern and a pattern detection separator circuit for outputting separately as a detection signal, by the identification pattern signal, an image input device being characterized in that so as to replace them with a black background to white background. 5. A character recognition unit interconnected via a system bus, an image input device and a microprocessor for controlling the whole, the image input device one-dimensionally converts image data from a target form. An image sensor to be read; an AD converter for digitizing an image signal read by the image sensor; and a normalized signal obtained by performing logarithmic conversion processing, shading processing, and the like on the digitized image signal of the form. , A binarization circuit that binarizes the normalized signal and outputs it as a character binarized signal, a pattern generator that generates a specific identification pattern signal, and the normalized signal that is input. A first binarization circuit for detecting dark black with a higher threshold (darker density) than the character binarization circuit for generating the character binarization signal; A second binarization circuit for background detection capable of switching between two thresholds, a lower threshold (lighter density) and a higher threshold than the character binarization circuit, and an output of the first binarization circuit. It has a first two-dimensional information calculating means for performing black density determination in units of n × n pixels, and two functions for performing white density determination in units of n × n pixels for the output of the second binarization circuit. A second two-dimensional information calculating means for detecting white of the background and outputting it as a white detection signal, wherein the threshold value of the second binarization circuit and the threshold value of the second binarizing circuit are determined by the result of the first two-dimensional information calculating means. A background detection circuit that performs an operation of switching the two functions of the second two-dimensional information calculation means; and a background detection signal composed of the white detection signal is input, and when n and m are arbitrary positive integers, n × the background information output in units of n pixels in one block. As a unit, m
A fourth two-dimensional information calculating means having two functions for performing a determination operation using the number of different white blocks as a threshold value for each of the × m block planes, and a function of the four functions of the fourth two-dimensional information calculating means. A second switching means for selecting and outputting one of the outputs, and performing an S (S is a positive integer) number of different two-dimensional information calculations on the output of the second switching means in two ways on an m × m block plane. Fifth with a plurality of arithmetic functions capable of
A dimension information calculation unit, a character obtained from an area determination result as a final determination result after S times of the fifth two-dimensional information calculation unit, and two threshold value determination results of the fourth two-dimensional information calculation unit. The photograph area determination results are sequentially stored, and the current m
Using the character / photo area determination result on the × m block plane and the character / photo area determination result obtained from the area around the current m × m block plane and stored, A global area determining means for globally determining a character / photograph area from a background wider than a block plane of × m, and the second switching means and the fifth and second elements are determined by the determination result of the global area determining means. An area determination circuit that switches the plurality of arithmetic functions of the dimension information calculation means and outputs the area determination result to the outside as an area switch signal; and based on the area switch signal, the character binary signal and the An area switching circuit for selecting and outputting any one of the identification pattern signals; and a character binarization signal discriminating a character binary signal and an identification pattern signal coming from the area switching circuit. And a pattern detection / separation circuit that individually outputs the pattern detection signal and a pattern detection signal, wherein the microprocessor detects a pattern output from the image input device when detecting a character frame that is not a dropout color printed thick on the form. Based on the signal, the cutout position of the target character written in the character frame is known, and a character recognition unit control signal instructs the character recognition unit to cut out the character. The character binarization signal is fetched in response to the input of a unit control signal, and character cutout detection is performed, thereby selectively detecting and removing the character frame that is not a thickly printed dropout color. Optical character recognition device.