JPH039487A - Processor and image character/picture deciding device - Google Patents

Abstract

PURPOSE:To surely remove paper noise having almost the same level as that of a character by dividing a character into a character area and a non-character area, outputting a character binarizing signal in the case of the character area, or outputting a white picture element signal in the case of the non-character area. CONSTITUTION:A matrix arithmetic circuit M1 outputs '1' (white) when the center picture element of a matrix consisting of 3 X 3 picture elements e.g. is '1' (white) or all the picture elements surrounding the center picture element are '1' (white). A matrix arithmetic circuit M2 executes black picture element blocking processing, a matrix arithmetic circuit M3 executes black density deciding processing and a matrix arithmetic circuit M4 executes character effective area deciding processing. A matrix arithmetic circuit M5 executes correction processing. A selector 2b in an output part 2 is operated through a character area deciding signal 6, and in the case of the character area, a character binarizing signal 5a is outputted after picture element unit noise processing. In the case of the non-character area, white data are outputted as a character binarizing signal 5b after noise processing. Consequently, comparatively large noise similar to the size of each character size can be surely removed.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to image processing technology and text/photo recognition technology, and is particularly effective in discriminating relatively large paper surface noise when reading text information from a form. Regarding technology.

[Conventional technology]

For example, in electronic file systems and the like, character information written on a form or the like is input as a character binary signal of a predetermined resolution using an image scanner or the like.

By the way, when reading text information on such forms, it is important to effectively remove paper surface noise other than the desired text information caused by dirt on the paper surface, etc., from the perspective of improving the quality of input data. becomes important.

Conventionally, as a technique for removing such paper noise, when targeting random noise that exists independently for each pixel, image smoothing and averaging processing are effective in suppressing noise. It has been known. Furthermore, when dealing with monochrome speckled noise (salt-and-pepper noise), which is a type of random noise, it is known that filtering processing, typified by a median filter, is effective for noise removal.

For details of the above-mentioned prior art, see Kyoritsu Shuppan Co., Ltd., May 25, 1985, first edition No. 3.
Published by Hiroshi Ozaki, Keiji Taniguchi, and Hideo Ogawa, “Image Processing JP
145 et seq., et al.

[Problem to be solved by the invention]

However, the above-mentioned conventional technology performs noise processing on a pixel-by-pixel basis by comparing the pixel of interest with neighboring pixels, so it is not effective in removing noise when the noise size is close to that of the target character image. The problem was that it could not be done.

SUMMARY OF THE INVENTION Therefore, an object of the present invention is to provide an image processing technique that can separate relatively large paper noise and character information.

Another object of the present invention is to provide an image processing device that can remove relatively large page noise with a simple circuit configuration.

Still another object of the present invention is to provide a text/photo identification device that has a text document reading mode and is capable of separating relatively large paper surface noise from text information.

A brief overview of typical inventions disclosed in this application is as follows.

[Means to solve the problem]

A brief overview of typical inventions disclosed in this application is as follows.

That is, the image processing apparatus according to the present invention includes an input means for generating a character binary signal and a noise processing means for removing paper noise from the character binary signal. a pixel-by-pixel noise removal processing means that performs two-dimensional information calculation on n×n pixels and removes noise on a pixel-by-pixel basis; A block-by-block black pixel detection means that performs two-dimensional information calculation to detect a set of black pixels in blocks of m×m pixels, and a block-by-block black pixel detection output obtained from the block-by-block black pixel detection means. Character area determination means performs eight different two-dimensional information calculations on a k×k block plane and determines a character area in block units, and when this character area determination means determines a character area, it performs eight different two-dimensional information calculations on a block plane of k×k. The apparatus is provided with a switching means for outputting a character binary signal obtained by performing noise removal processing, and outputting a white signal if not.

Further, the text/photo identification device of the present invention is a text/photo identification device that discriminates between text areas and photo areas based on background white information in pixel units and block units. A mode signal and a first switching means that performs a pixel-by-pixel noise removal process by switching the calculation function of the first two-dimensional information calculation means that performs a pixel-by-pixel character/photo determination process based on the character document reading mode signal. Then, in response to the character original reading mode signal, the input signal to the second two-dimensional information calculation means for determining white density is switched from the background white signal to the character binary signal, and a black block containing a set of black pixels is detected. and a third two-dimensional information calculation means for determining whether a character is a background noise or a character by switching a calculation function of a third two-dimensional information calculation means for determining characters/photos in units of blocks based on a character document reading mode signal. 3 switching means are provided.

[Effect]

According to the above-described image processing device of the present invention, a character binary signal from which noise has been removed pixel by pixel is used to determine whether or not a certain amount of large black pixels (for example, 2×2 pixels) are included, that is, whether or not the block is a black block. in desired block units (e.g. 8
x 8 pixel block), and then based on the judgment result for each block, black pixels are detected in a large range by thresholding the number of black blocks on an m x m (for example, 4 x 4) block plane. The black density determination to be performed and the result of the black density determination are divided into larger blocks (for example, 1 block = 16
Character area determination that determines whether there is a large surrounding area where black pixels can be found by considering the center block as a black block or white pixel block depending on the presence or absence of black blocks in the four corner blocks. Then, the result of this character area determination is expanded to a desired number of blocks, and if the number of black blocks in the block exceeds a predetermined threshold, it is treated as a character area, otherwise it is removed as a collection process. By performing three different two-dimensional information operations consisting of By outputting a white pixel signal, it is possible to accurately remove not only pixel-by-pixel noise but also paper surface noise of the same size as characters.

Further, by dividing a plurality of two-dimensional information calculations into line units and performing them in one matrix calculation circuit, relatively large paper noise can be removed with a simple circuit configuration.

Further, according to the above-described character/photo identification device of the present invention,
A first two-dimensional information calculation means that performs character/photo judgment processing in pixel units according to the setting of an external text document reading mode signal, a second two-dimensional information calculation means that performs background white density judgment, and block units. The third step is to judge the text and photo of
The calculation function of the two-dimensional information calculation means of the first two-dimensional information calculation means is switched.
Dimensional information calculation is performed by pixel-by-pixel noise processing, and the second
In the second two-dimensional information calculation, the input signal is switched from a background white signal to a character binary signal to detect a black block, and in the third two-dimensional information calculation, processing is performed to determine whether or not it is a character area. By doing so, in the character document reading process, not only noise in pixel units but also paper surface noise of the same size as the characters can be accurately removed.

[Embodiment 1] Hereinafter, an image processing apparatus that is an embodiment of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing an example of the configuration of an image processing device that is an embodiment of the present invention, FIGS. 2 and 3 are explanatory diagrams showing examples of the functions of each part, and FIG. 5 is an explanatory diagram showing an example of block settings in two-dimensional information calculation, and FIGS. 6 and 7 are block diagrams showing an example of a circuit configuration for realizing two-dimensional information calculation.

The image processing apparatus of this embodiment is broadly divided into a pixel-by-pixel noise removal processing section 1, an output section 2, a block-by-block black pixel detection section 3, and a block-by-block character area determination section 4.

The pixel-by-pixel noise removal processing unit 1 performs pixel-by-pixel noise removal processing (isolated noise removal processing) on the character binary signal 5 that arrives from an image input device (not shown), and converts the character binary signal after the pixel-by-pixel noise processing to the character binary signal 5. It is provided with a matrix calculation circuit M1 that performs an operation of outputting a converted signal 5a.

That is, as shown in FIG. 2, this matrix calculation circuit M1 outputs 1 (white) when the central pixel of a matrix consisting of 3×3 pixels is 1 (white) or all surrounding pixels are 1 (white). ).

The pixel-by-pixel noise-processed character binary signal 5a output from the pixel-by-pixel noise removal processing unit 1 is sent to a subsequent output unit 2.
and the block-by-block black pixel detection section 3 is manually operated.

The block unit black pixel detection section 3 includes a matrix calculation circuit M2, and this matrix calculation circuit M2
As shown in FIG. 2, based on the incoming pixel-by-pixel noise-processed character binary signal 5a, 2×2 pixels are divided into 8×8 pixels.
If there is one or more black pixels, 1 is output, otherwise O is output to the subsequent block unit character area determination unit 4 (black pixel blocking process).

Further, a block-by-block character area determining section 4 which receives the output from the block-by-block black pixel detecting section 3 includes a plurality of matrix calculation circuits M3. M) IJx calculation circuit M
4. It is equipped with a matrix calculation circuit M5.

Each of the matrix calculation circuits M3 to M5 is designed to perform processing as shown in FIG.

That is, the matrix calculation circuit M3 calculates, for example, 5 pixels in a 4×4 block each consisting of 8×8 pixels.
A process (black density determination process) is performed to output 1 when there is a black block or more, and 0 otherwise, and the matrix calculation circuit M4 has a size of 16 x 16 pixels each, as shown in FIG. The matrix for the black density determination process is expanded to the four corners of the 4×4 block, and if even one of the four corner blocks is 1, 1 is output, otherwise 0 is output (character effective area determination process). .

Furthermore, as shown in FIG. 5, the matrix arithmetic circuit M5 includes, for example, 6 black blocks in which the output from the preceding stage matrix arithmetic circuit M4 is 1 in 3×3 blocks each having a size of 16×16 pixels. 1 if there are blocks or more
(i.e. character area), otherwise 0 (non-character area)
is sent to the output unit 2 as a character area determination signal 6 (collection determination).

The output section 2 processes the pixel-by-pixel noise-processed character binary signal 5a coming from the pixel-by-pixel noise removal processing section 1 at the previous stage, by subjecting it to the above-described processing in the block-by-block black pixel detection section 3 and the block-by-block character area determination section 4. Line memory 2a whose capacity is set to store an amount corresponding to the time required for
Then, the character binary signal 5a after pixel unit noise processing coming from this line memory 2a and predetermined white data (white=
1) is selected according to the character area determination signal 6 inputted in synchronization with the pixel unit noise-processed character binary signal 5a, and outputted as the noise-processed character binary signal 5b. It is composed of a selector 2b.

As a result, when the pixel-by-pixel noise-processed character binary signal 5a that arrives via the line memory 2a is determined to be paper surface noise by the block-by-block black pixel detection unit 3 and the block-by-block character area determination unit 4, Paper surface noise is removed by suppressing the output of the pixel unit noise-processed character binary signal 5a and outputting white data as the noise-processed character binary signal 5b.

Here, FIG. 6 shows an example of the configuration of the above-mentioned matrix calculation circuit Ml-=M5.

That is, the matrix calculation circuit shown in the same figure has 5
It performs two-dimensional information calculations with a matrix size of It is comprised of a memory C dedicated to continuous processing for calculations for performing desired two-dimensional information calculations. Further, the calculation read-only memory C can switch the calculation function by a function switching signal, and can perform three two-dimensional information calculations at the same time.

By the way, multiple matrix operations in block units are
It only needs to be performed once per block, and if the block size is n pixels x n lines, one calculation can be performed once per n lines. It is possible to perform n two-dimensional information operations with one matrix operation circuit (however, there are multiple operation outputs).

Therefore, in the case of this embodiment, the functions of the matrix calculation circuits M3 to M5 constituting the block unit character area determination section are performed by one matrix calculation circuit M as shown in FIG. Efforts are being made to reduce the circuit scale.

That is, the block unit character area determination unit 4 of this embodiment includes the matrix calculation circuits M3. M4. A matrix calculation circuit M that simultaneously performs and outputs three calculations to be performed by M5, and a selector 4b and a selector 4C that operate according to a control signal from a timing control circuit 4a.
A line memory 4d holds the result of matrix calculation for one line and uses it as input for the next calculation process, and holds and outputs the final calculation result for two lines of n lines, which is output once every n lines. It is composed of a line memory 4e.

Hereinafter, an example of the operation of the image processing apparatus of this embodiment will be described with reference to flowcharts in FIGS. 8 and 9, as well as in FIGS. 10 and 11.

First, a character binary signal 5 arriving from an image input device (not shown) is input to the pixel-by-pixel noise removal processing section 1, and is processed into one bit (one pixel) by matrix calculation as shown in FIG. 2 in the matrix calculation circuit M1. ) is checked for the presence of isolated noise (step 1001),
If so, it is removed (step 1002).

Note that the above processing also serves as preprocessing of black pixels in blocks, which will be described later.

In this way, the pixel-by-pixel noise-processed character binary signal 5a from which pixel-by-pixel noise has been removed is held in the line memory 2a of the subsequent output section 2, and at the same time, part of it is stored in the block-by-block black pixel detection section 3. It is manually input to the matrix calculation circuit M2.

This matrix arithmetic circuit M2 determines whether there is a set of 2X2 black pixels in an 8X8 pixel block (step 1003), and if so, the block is designated as a black pixel block (step 1004); otherwise, it is a white pixel block. (step 1005), black pixel blocking processing is performed.

In other words, in an area with text, a collection of 2x2 black pixels can be easily found, but in areas away from the text, pixels smaller than 2x2 pixels may occur frequently as paper noise (in the case of this example). , this noise is removed in the previous stage matrix calculation circuit M1), 2
Noises larger than X2 pixels rarely occur, and even if they do occur, they occur only once and in a discrete manner, so by this black pixel blocking process, it is possible to extract those that are likely to be noises in areas other than character areas.

Further, in the matrix calculation circuit M3 (M) of the block unit character area determination section 4, for example, assuming that the size of paper surface noise to be removed is less than 1 tap angle, the 0, 5M square ( If the line density is 16 lines/rmr, a black pixel determination block with a size of 8 x 8 pixel blocks) is developed on a 4 x 4 block plane (step 100).
6), Determine whether the number of black pixel blocks is 5 or more or less (step 1007), if it is 5 or more, treat it as a black pixel block as a character (step 1008), and if it is 4 or less, consider it as noise. A white pixel block is output (step 1009>Processing (black density determination processing) is performed. As a result, isolated noise having a size of 1fi angle or less can be determined to be noise.

By the way, in this black density determination process, we are only looking at pixels within a 2 mm square area, so for example, periods (,
) or comma (1), a collection of small black dots (characters) close to a character may be considered noise.

Therefore, in the matrix arithmetic circuit M4 (M), the judgment unit is 4×4 of l mm square as shown in FIG.
7) A character effective area determination matrix for IJ hook size is set, and the output of the 2ffi11 square black density determination matrix in the black density determination process is expanded into this matrix (step 1010).

In this character effective area determination matrix,
It is checked whether at least one of the four corner elements is a black pixel block (step 1011), and if so, it is determined that it is a character area and a black pixel block is output (step 1012); if not, it is regarded as noise. Character valid area determination processing is performed to output a white pixel block (step 1013).

Note that the reason why the center block is not looked at in the character valid area determination matrix is that even if there is a block in which black pixel blocks are detected in the center block, if all surrounding blocks do not detect black pixel blocks, This is because the center block is regarded as noise.

Furthermore, as shown in FIG. 5, the matrix calculation circuit M5 (M) calculates 3
Expand the collection determination matrix for x3 blocks (step 1014). For example, if there is a block (black pixel block) that has been determined to be a valid character area for 6 blocks or more, set it to 1, and finally define it as a character area (step 1015).
, otherwise it is set as outside the character area (step 10
16) Performs a correction process to output the character area determination signal 6.

With this process, when setting the block size in the series of processes as described above, it is possible to smooth the effective character area with a width of 7 squares.

In other words, this correction process is a so-called smoothing process, and since the judgment in the previous stage of character valid area judgment is performed by focusing only on the four corner blocks, the detection unit is discontinuous, and if it is left as it is, the character is not valid. Since the region determination results are also discontinuous, this collection process is used to ensure continuity in the final determination results.

Then, the selector 2b of the output unit 2 is operated via the character area determination signal 6, and in the case of a character area, the character binarized signal 5a after pixel unit noise processing is output (step 1018), and in the case of a non-character area, the character binarized signal 5a is output. White data (step 1019)
, and output as a character binary signal 5b after noise processing.

This makes it possible to accurately remove relatively large page noise that is close to the size of characters outside the character area.

As a result, the quality of input images such as text manuscripts is improved.

Furthermore, the three types of matrix operations to be performed in the block unit character area determination section 4 are performed by one matrix operation circuit M in a time-sharing process using the circuit configuration shown in FIG. It is possible to reliably prevent the circuit scale from increasing unnecessarily due to the noise removal process, and it is possible to suppress an increase in cost when manufacturing an image processing apparatus having a paper noise removal function as in this embodiment.

[Embodiment 2] Next, an example of a case where a noise removal process is performed in a text/photo identification device having a function of cutting out a text image from a document containing a mixture of texts and pictures will be described with reference to FIG. 12. .

The details of the operation of each part in the text/photo identification device of this embodiment in the reading process of a mixed text/photo document can be found in Japanese Patent Application No. 1983, which has already been filed by the applicant.
Since it is detailed in the specification of No. 307165, in the following explanation, the explanation of the configuration and operation of the present invention other than the part related to the noise removal process will be simplified.

The text/photo identification device of this embodiment is broadly divided into a background detection circuit 100, a text binary signal processing circuit 200, an area judgment circuit 300, and an area switching circuit 400.

First, the background detection circuit 100 will be explained.

The binarization circuit 51 receives a slice signal 10 and a video normalized signal 11 coming from an image input device (not shown) as input, detects deep black, and outputs the matrix calculation circuit 6 at the subsequent stage.
Output to 1. This matrix calculation circuit 61 performs black density determination and outputs a black detection signal 14. The binarization circuit 51 generates a slice signal 10 and a video normalized signal 11.
The white background is detected from the black detection signal 14 and outputted to the matrix calculation circuit 62. This matrix calculation circuit 6
2 performs white density determination using a function instructed by the black detection signal 14, and outputs a width detection signal 15.

Next, the character binary signal processing circuit 200 will be explained.

The matrix calculation circuit 63 manually determines the pixel positions inside, outside, and around the black image using the character binary signal 12, and outputs the determined pixel positions to the selector control circuit 71. In addition, the selector control circuit 7
1 is the determination result of the inside, outside, and periphery of the black image and the black detection signal 1
4 and the width detection signal 15, the text/photo area is determined in pixel units, and the determination signal is output to the selector 81. The selector 81 manually inputs the character binary signal 12 and the pseudo halftone signal 13 and outputs one of the signals for each pixel based on character/photo area determination for each pixel.

Further, the configuration of the area determination circuit 300 is as follows.

The matrix calculation circuit 64 inputs the width detection signal 15 and further performs density determination (
background density determination). The background density determination is performed for two functions, and the respective determination results are output to the selector control circuit 72 and the selector 82. selector 82
The processing results of the two functions are sent to the subsequent matrix calculation circuit 65 according to the output of the selector control circuit 72.
Switch whether to output to 1. Matrix calculation circuit 6
51 performs binary judgment (character area judgment) processing that performs thick processing for background density judgment, and prints the judgment result.
It is output to the J-Fuchs calculation circuit 652. The matrix calculation circuit 652 and the matrix calculation circuit 653 perform correction processing on the binary determination results. That is, the matrix calculation circuit 652 right and the matrix calculation circuit 653 each have two functions, and can be switched to either one by the output from the selector control circuit 72. Matrix calculation circuit 65
3 outputs an area switching signal 16 as the final determination result. Further, the selector control circuit 72 performs global character/photo judgment based on the outputs of the two functions of the matrix calculation circuit 64 in the previous stage and the area switching signal 16, and performs global character/photo judgment using the outputs of the two functions of the matrix calculation circuit 64 in the previous stage and the function of the selector 82 and matrix calculation circuits 652 and 653. Reflect in switching.

Next, the area switching circuit 400 will be explained.

Line memory 91 and line memory 92 operate to delay pseudo halftone signal 13 and character binary signal 12 by an amount commensurate with the determination signal processed by background detection circuit 100 and area determination circuit 300. Although the pseudo halftone signal 13 is input to the line memory 91, the line memory 92 is not input with the character binary signal 12 itself;
When a photographic area is determined pixel by pixel by region determination, the selector 81 manually inputs a signal replaced with a pseudo halftone signal. The selector 83 selects one of the outputs of the line memory 91 and the line memory 92 in response to the area switching signal 16 and outputs it as an adaptive binary signal 17.

Due to the structure of each part, normal text/photo recognition processing is difficult to detect when the white level of the background is close to the threshold for character binarization, such as a form with a mixture of text and photos, or when densely packed characters such as kanji are processed. This makes it possible to accurately extract characters even in cases where the internal background is higher than the threshold for character binarization.

Here, in this embodiment, the following selector 84 and AND gate 9 is installed in the character/photo judgment device configured as described above.
1 and text document reading mode signal 1, which is set as appropriate by the user when not using a text/photo mixed document.
8 is added to perform the relatively large page noise removal process in the character document input process, as in the case of the first embodiment.

In addition, in FIG. 12, the portion added to realize the noise removal function is shown by a thick line.

That is, the character original reading mode signal 18 is transmitted to the matrix calculation circuit 63. Matrix calculation circuit 64, matrix calculation circuit 652. The numerical arithmetic operations performed in each of the matrix arithmetic circuits 653 are switched to functions for noise removal processing.

A selector 84 interposed between the binarization circuit 52 and the matrix calculation circuit 62 constituting the background detection circuit 100 selects the input to the matrix calculation circuit 62 when the character document reading mode signal 18 is set. An operation is performed to switch from the output from the binary circuit at the time to the output from the matrix calculation circuit 63 of the character binary signal processing circuit 200.

Also, the pseudo halftone signal 1 input to the line memory 91
The AND gate 91 provided in the path No. 3 is configured to always output white data at the "0" level according to the setting of the character original reading mode signal 18.

Then, when setting the text document reading mode signal 18, the matrix calculation circuit 63 removes isolated point noise,
The matrix calculation circuit 62 performs black pixel blocking processing (however, the function is the same as in the case of character/photo recognition), the matrix calculation circuit 64 performs black density determination processing, and the matrix calculation circuit 651 performs motion area determination processing for characters. Deeds (
However, the function is the same as in the case of text/photo recognition), the matrix calculation circuit 652 performs correction processing, and the matrix calculation circuit 652 performs correction processing.
+J Hooks calculation circuit 653 performs the same output as input without doing anything, and outputs character area determination signal 6a to selector 83. The selector 83 selects from the matrix calculation circuit 63,
The pixel unit noise-processed character binary ratio signal 5a held in the line memory 92 via the selector 81 is output as the noise-processed character binary ratio signal 5b.

As a result, the text/photo identification device of this embodiment can not only read originals containing text and photos in normal operation, but also enter a mode dedicated to input processing of text originals by setting the text original reading mode signal 18. by doing,
It only removes noise on a pixel basis.Input processing can be performed that can accurately remove relatively large paper noise that is close to the size of characters.

Although the invention made by the present inventor has been specifically explained above based on Examples, it goes without saying that the present invention is not limited to the Examples and can be modified in various ways without departing from the gist thereof. Nor.

For example, the threshold values used for determination on a block-by-block basis at each stage are not limited to those exemplified in the above embodiments.

Further, the configuration of the matrix calculation circuit that performs desired two-dimensional information calculation is not limited to that exemplified in the above embodiments.

〔Effect of the invention〕

Among the inventions disclosed in this application, the effects obtained by typical inventions are briefly described below.

That is, an image processing apparatus according to the present invention includes an input means for generating a character binary signal, and a noise processing means for removing paper surface noise from the character binary signal. pixel-by-pixel noise removal processing means that performs two-dimensional information calculation of n×n pixels on the encoded signal and removes noise on a pixel-by-pixel basis; block-by-block black pixel detection means that performs two-dimensional information calculation on m pixels and detects a set of black pixels in m×m pixel block units; and the block-by-block black pixels obtained from this block-by-block black pixel detection means. Character area determination means performs five different two-dimensional information calculations on the kxk block plane on the detection output and determines the character area in block units, and if the character area determination means determines that it is a character area, the pixel A switching means is provided for outputting the character binary signal obtained by performing unit noise removal processing, and outputting a white signal if not, so that the character 2 from which noise has been removed pixel by pixel is provided.
Using the value signal, a certain amount of large black pixel (for example, 2
x2 pixels), that is, whether it is a black block or not, in a desired block unit (for example, a block of 8 x 8 pixels).
Then, m×m is determined based on the determination result for each block.
(for example, 4x4) block plane, black density judgment detects black pixels in a large range by threshold judgment of the number of black blocks, and the result of the black density judgment is applied to a larger block (for example, 1 block - 16 x 16 For example, depending on the presence or absence of black blocks in the four corner blocks, the center block is regarded as a black block or a white pixel block, and it is determined whether there is a large surrounding area where black pixels can be found. , the result of this character area determination is expanded to a desired number of blocks, and if the number of black blocks in the block exceeds a predetermined threshold, it is treated as a character area, otherwise it is removed as a collection process. By performing 8 different two-dimensional information calculations, it is divided into areas that are close to characters and areas that are not. In the case of character areas, a character binary signal is output, and in the case of non-character areas, a character binary signal is output. By outputting the white pixel signal, it is possible to accurately remove not only noise in pixel units but also paper noise of the same size as characters.

Further, by dividing a plurality of two-dimensional information calculations into line units and performing them in one matrix calculation circuit, relatively large paper noise can be removed with a simple circuit configuration.

Further, according to the text/photo identification device of the present invention, the text/photo identification device discriminates between a text area and a photo area based on background white information on a pixel-by-pixel and block-by-block basis.
A character original reading mode signal dedicated to character originals and a calculation function of the first two-dimensional information calculation means that performs character/photo determination processing on a pixel basis are switched based on this character original reading mode signal, and noise removal processing is performed on a pixel-by-pixel basis. and a first switching means for switching the input signal to the second two-dimensional information calculation means for determining white density from the background white signal to the character binary signal according to the character document reading mode signal, A second switching means detects a black block containing a set of , and a third two-dimensional information calculating means performs character/photo judgment in block units according to the text document reading mode signal. Since the first two-dimensional information is provided with a third switching means for determining whether it is noise or text, the first two-dimensional information is subjected to pixel-by-pixel text/photo identification processing by setting an external text document reading mode signal. Switching the calculation functions of the calculation means, the second two-dimensional information calculation means for determining the white density of the background, and the third two-dimensional information calculation means for determining the characters/photos in units of blocks, and performing the first two-dimensional information calculation. In the second two-dimensional information calculation, the input signal is switched from the background white signal to the character binary signal to detect black blocks, and in the third two-dimensional information calculation,
By having the dimensional information calculation perform processing to determine whether or not it is a character area, it is possible to accurately remove not only pixel-by-pixel noise but also paper surface noise of the same size as the characters in the text document reading process. be able to.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an example of the configuration of an image processing device that is an embodiment of the present invention, FIG. 2 is an explanatory diagram showing an example of the functions of each part, and FIG. FIG. 4 is an explanatory diagram showing an example of a function; FIG. 4 is a diagram showing an example of a matrix setting method in two-dimensional information calculation; FIG. 5 is a diagram showing an example of a matrix setting method in two-dimensional information calculation; FIG. 6 is a block diagram showing an example of a circuit configuration for realizing the function, FIG. 7 is a block diagram showing an example of a circuit configuration for realizing the function, and FIG. 8 is a block diagram showing an example of the state of the read document. FIG. 9 is an explanatory diagram illustrating threshold setting in the relationship between a black pixel set and a black pixel determination block, and FIG. 10 is a flowchart illustrating an example of noise removal processing in the image processing apparatus of the present invention. FIG. 11 is a flow chart showing an example of noise removal processing in the image processing device of the present invention, and FIG. 12 is a block diagram showing an example of the configuration of the text/photo determining device of the present invention. 1... Pixel unit noise removal processing section, 2... Output section,
2a...Line memory, 2b...Selector, 3...
- Block unit black pixel detection unit, 4... Block unit character area determination unit, 4a... Timing control circuit, 4b,
4c...Selector, 4d, 4e...Line memory,
5... Character binary signal, 5a... Character binary signal after pixel unit noise processing, 5b... Character binary signal after noise processing, 5b... Character binary signal after noise processing ,6.
6a... Character area determination signal, 10... Slice signal, 11... Video normalized signal, 12... Character binary signal, 13... Pseudo halftone signal, 14... Black detection Signal, 15... Width detection signal, 16... Area switching signal,
17...Adaptive binary conversion signal, 18...Character document reading mode signal, 51.52...Binarization circuit, 61, 62
,63゜64...Matrix calculation circuit, 651,6
52.653...Matrix calculation circuit, 71...
Selector control circuit, 72...Selector control circuit, 81
... selector, 82... selector, 83... selector, 84... selector, 91... AND game),
91.92...Line memory, 100...Background detection circuit, 200...Character binary signal processing circuit, 300
... area determination circuit, 400... area switching circuit, M,
Ml, M2 M3. M4. M5...Matrix calculation circuit, AI...Register, B...Line memory, C
...Memory dedicated to continuous operation for calculation, D...Function switching signal,
1001 to 1019...Steps showing an example of noise removal processing.

Claims (1)

[Claims] 1. Input means for generating a character binary signal; and the character 2
In an image processing device comprising a noise processing means for removing paper noise from the digitized signal, the character binary signal is
a pixel-by-pixel noise removal processing means that performs two-dimensional information calculation of ×n pixels and removes noise in pixel units;
block-by-block black pixel detection means that performs two-dimensional information calculation of ×m pixels and detects a set of black pixels in block units of an m×m image; character area determination means for performing S different two-dimensional information calculations on pixel detection outputs on a k×k block plane and determining character areas on a block-by-block basis;
If the character area determination means determines that the character area is a character area, the character binary signal obtained by performing pixel-by-pixel noise removal processing is outputted, and if not, the switching means outputs a white signal. An image processing device comprising: 2. The image processing apparatus according to claim 1, wherein the plurality of two-dimensional information calculations are performed by one matrix calculation circuit by dividing the two-dimensional information into line units. 3. In a text/photo identification device that discriminates text areas and photo areas based on background white information in pixel units and block units, a text original reading mode signal dedicated to text originals;
A first switching means that performs a pixel-by-pixel noise removal process by switching the arithmetic function of the first two-dimensional information calculation means that performs a pixel-by-pixel character/photo determination process in response to the character document reading mode signal; The input signal to the second two-dimensional information calculation means for determining white density is switched from the background white signal to the character binary signal in response to the document reading mode signal, and the black block including the set of black pixels is detected. A third switching means for determining whether background noise or characters is background noise or characters by switching the calculation function of the third two-dimensional information calculation means for determining characters/photos in blocks based on the switching means 2 and the character original reading mode signal. What is claimed is: 1. A text/photo identification device comprising: a switching means;