JPH06141176A - Picture discrimination device - Google Patents

Abstract

PURPOSE:To improve the accuracy of photograph/character discrimination without enlarging a reference area. CONSTITUTION:The pixel data of the 4X4 reference area including a noticeable picture element are outputted from a storage part 1. A correction table part 2 counts the number A of the picture element value variation points of the first line of the reference area and the picture element value variation points between adjacent lines and the number B of the picture element value variation points of a first column and the picture element value variation points between the adjacent columns and outputs a='1' when both A and B are equal to or more than a stipulated value. An area discrimination part 3 counts the consecutive appearance number of the picture element of a='1' and when a counted value is equal to or more than the stipulated value, discriminates photograph pictures and outputs b='1'. The correction table part 2 selects smoothing correction for the photograph pictures at the time of b='1' and selects the smoothing correction for character pictures at the time of b='0'.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image discriminating apparatus for discriminating whether binary image data is a photographic image or a character image.

[0002]

2. Description of the Related Art In a facsimile machine, 200 × 2
When converting the received data of 00 dpi into a recording pixel size of 400 × 400 dpi, a method of performing smoothing correction for smoothing the outline of a character is known (Japanese Patent Laid-Open Nos. 3-82267 and 3-83267). 822
No. 68).

Even if this smoothing correction of characters is effective when applied to a character image, if it is applied to a photographic image, the gradation is lost and the image quality deteriorates. The cause is that when the image is doubled in the main scanning direction and the sub-scanning direction, it may be enlarged by a signal different from the received data in order to smooth the contour of the character. However, since the density of binary image data of a photographic image is expressed by area modulation by a dither method or an error diffusion method, the density of a single area changes in smoothing correction of characters.

Therefore, conventionally, there is also known a method of discriminating whether the received data is a character image or a photographic image and automatically controlling whether or not the smoothing correction is performed according to the result. For example, in Japanese Unexamined Patent Publication No. 3-254276, a reference area of 6 (sub-scanning direction) × 9 (main-scanning direction) pixels is used to determine the spatial frequency of the received image, the periodicity, and the majority of the determination results based on the pixel isolation. A method is described in which image areas of a character area and a photograph area are separated and smoothing correction is performed only on the character area.

The distinction between a character image and a photographic image is a technique required not only in the field of facsimile described above but also in various fields of handling image data.

[0006]

Generally, the more pixels that are referred to in order to distinguish between a character image and a photographic image, the more accurate the discrimination becomes. Regarding this, FIG. 9 and FIG.
This will be described with reference to 0.

FIG. 9 shows a pseudo halftone representation by the dither method. Here, it is assumed that a uniform density region of 8 × 8 pixels is subjected to dither processing by using a 4 × 4 threshold matrix having a threshold arrangement as shown in the upper left of (a). If each pixel has a density equal to or higher than the value at the corresponding position of the threshold matrix, it is expressed in black, so that the area of density 0 (white) is as shown in (a).
The area of density 1 is shown in (b), and the area of density 2 is (c).
like,. ． ． , The area of density 15 is (q),
The density is expressed by the area ratio of black and white, respectively. Although not shown, all pixels in the black area are black pixels.

FIG. 10A shows an example of a character image.
9B shows an example of a dithered photographic image, which corresponds to the pattern of the area of density 8 shown in FIG. 9I.

With respect to the character image and the photographic image, it is assumed that the character / photo is discriminated by referring to the area of the 4 × 4 matrix including the target pixel. For example, focusing on the pixels marked with a star, the pattern of the 4 × 4 reference area is
The character image looks like (c), and the photographic image looks like (d), and both patterns are the same. That is, in the 4 × 4 reference area, character / photo discrimination cannot be performed.

In this example, the reference area is 4 (in the sub-scanning direction).
By enlarging to a matrix of × 8 (main scanning direction), the character image and the photographic image have different reference area patterns as shown in (e) and (f), and therefore character / photo discrimination is possible. Becomes

By thus increasing the reference area, the accuracy of character / photo discrimination can be improved. Therefore, even in the example shown in the above-mentioned Japanese Patent Laid-Open No. 3-254276, 6
The reference area of × 9 (54 pixels) is used.
However, when the number of reference pixels increases, there is a problem in that the circuit for character / photo identification becomes significantly complicated and expensive.

The present invention has been made in view of the above problems, and an object thereof is to provide an image discriminating apparatus for discriminating a character image from a photographic image with high accuracy without increasing the reference area. .

[0013]

In order to solve the above-mentioned problems, the first invention relates to each pixel of binary image data,
A means for determining whether there are many changes in pixel values in the reference area with respect to a reference, and the number of consecutive pixels that have been determined to have many changes by this means are counted and calculated. And a means for determining that it is a photographic image when the numerical value is equal to or larger than a predetermined value, and determining that it is a character image otherwise.

According to a second aspect of the present invention, for each pixel of the binary image data, there is a judging means for judging whether or not there are a large number or a small number of changing points of the pixel value in the reference area with respect to a reference, and the judging means changes the value. A first counting unit that counts the number of consecutive pixels that are determined to have a large number of points, and a second counting unit that counts the number of consecutive pixels that are determined to have a small number of changed points by the determination unit. Of the first counting means and the count value of the first counting means is greater than or equal to a predetermined value and the count value of the second counting means is less than the predetermined value. A numerical value is less than a predetermined value and the second count value is greater than or equal to a predetermined value, and a character image is determined, and under other conditions, the same determination as that of the immediately preceding pixel is included. is there.

According to a third aspect of the present invention, with respect to each pixel of the binary image data, there is a determination means for determining whether or not there are more or less pixel value change points in the reference area with respect to a reference, and a predetermined number of consecutive pixels. Among the above, there is provided a structure for determining whether it is a photographic image or a character image according to the number of pixels determined by the determination means to have a large or small number of pixel value changes. is there.

[0016]

According to the first to third inventions, a photographic image and a character image are discriminated by using the judgment results of the number of pixel value change points for a plurality of consecutive pixels in an accumulated form. It can be considered that the size of the reference area is larger than the reference area of each pixel. Therefore, the size of the reference area should be minimized and the reference area size cannot be used for discrimination. It is possible to distinguish such an image.

As will be described in more detail with reference to the embodiments described later, in the binarized character image and photographic image, when the change in pixel value in the reference area is examined in the main scanning direction and the sub scanning direction, the photographic image shows , While pixels with a large number of pixel value change locations in both sub-scanning directions tend to appear consecutively, in a character image, it is recognized that such a pixel has a small number of consecutive pixels and a small number of consecutive pixels. .

According to the first aspect of the present invention, a region in which a predetermined number or more of pixels determined to have a large number of such pixel value change points continue is determined as a photographic image, and a region other than that is determined as a character image. .

In a photographic image area, pixels with few pixel value changes may locally appear in a continuous area of pixels with many pixel value changes. According to the first aspect, such a pixel may be determined to be a character image.

According to the second aspect of the invention, the image discrimination is performed based on both the continuous number of pixels with many pixel value changing points and the consecutive number of pixels with few pixel value changing points. Further, according to the third invention. , An image area is determined by the number of pixels having a large number or a small number of pixel value change points included in a predetermined continuous plurality of pixels, and therefore a pixel area having a small number of pixel value change points locally appearing in a photographic image area, or vice versa,
It is also possible to correctly discriminate a region of pixels in which a lot of pixel value changes locally appear in the character image region. Further, according to the third invention, it is possible to easily avoid the influence of noise, and it is possible to stably carry out image determination in a large unit (for example, one line unit).

[0021]

1 is a block diagram of a recording apparatus according to a first embodiment of the present invention. In this embodiment, 200 × 200 dpi image data such as facsimile reception data is converted into 4
This is an example in which the pixel density is converted into a recording pixel size of 00 × 400 dpi and recording is performed.

In FIG. 1, reference numeral 1 denotes a storage unit having a memory capacity for one page, which is used for storing image data transferred to a terminal 8 from an external circuit (not shown). Reference numeral 2 denotes a correction table unit including a ROM or the like, which is used for pixel density conversion and smoothing correction for characters or smoothing correction for photographs, and calculation for determining whether there are many pixel value change points in the reference area (hereinafter referred to as correlation calculation Call) and output the result as a decorrelation signal a. Reference numeral 3 denotes an area determination unit that determines whether the image is a character image or a photographic image by using the accumulated results of the correlation calculation of the correction table unit 2 and returns the determination result to the correction table unit 2 as a determination signal b. In this embodiment, the counter circuit operates as described later. Reference numeral 4 denotes a storage unit for storing image data (correction data) that has been subjected to pixel density conversion and smoothing correction by the correction table unit 2 and has a memory capacity for one page. A recording unit 5 records the image data stored in the storage unit 4 on recording paper at 400 × 400 dpi. Reference numeral 6 is a control unit for controlling the above-mentioned units.

The operation of the recording apparatus having the above structure will be described below. First, the control unit 6 outputs the clear signal CLR to clear the storage units 1 and 4 and the area determination unit (counter circuit) 3.

Next, the controller 6 outputs the clock SCLK. Image data SiN for one page is sequentially transferred to the terminal 8 from an external circuit (not shown) in synchronization with the clock SCLK. This image data is sequentially written in the storage unit 1 in synchronization with the clock SCLK.

When one page of image data is stored in the storage unit 1, the control unit 6 stops the transmission of the clock SCLK to stop the transfer / storage of the image data, and another clock RC.
Output LK. The storage unit 1 synchronizes with the clock RCLK and outputs data DAT (0, 0) of 16 pixels (reference pixels) of a 4 × 4 matrix (reference area) including a target pixel to be processed.
0) to DAT (3,3) are output. Correction table section 2
Receives the reference pixel data as input, performs pixel enlargement twice in the main scanning direction and the sub-scanning direction on the target pixel, and performs smoothing correction for a character image or a photographic image.
The correction data CDT0 to CDT3 of × 2 pixels are output, and the correlation calculation for the reference pixels is performed.

FIG. 2 is an explanatory diagram of the reference pixel and the correction pixel for the target pixel. In FIG. 2, DAT
(1, 1) is the data of the pixel of interest.

The value of the 2 × enlarged pixel of the target pixel is determined based on the reference pixel data of the 4 × 4 matrix. As for the type of smoothing correction, the smoothing correction for photographs is selected when the output signal b of the area determination unit 3 is “1” (photographic image), and the smoothing for characters is selected when b = “0” (character image). The correction is selected. The correction data is clock R
The data is sequentially written in the storage unit 4 in synchronization with CLK. Figure 3
Indicates the operation timing at this time.

Here, the contents of the correlation calculation in the correction table section 2 will be described. FIG. 4 is a flowchart thereof.

First, the counter COUNT used for processing
Zero-clear A, COUNTB, I, J (S
1). The counters I and J indicate relative coordinates of pixels in the reference area in the sub-scanning direction and the main scanning direction, and the coordinates of one pixel are indicated by (I, J).

Next, for the four pixels on the first line (J = 0), DAT (0,0), DAT (0,1), and DAT
(0,1) and DAT (0,2), DAT (0,2) and D
AT (0,3) are compared in order, and the counter COUNTA is incremented each time there is a mismatch (S2 to S).
5). That is, for the first line, the locations where the values of the pixels adjacent to each other in the main scanning direction are different (pixel value change locations) are counted.

Next, the counters I and J are cleared to zero (S
After 6), the number of changed pixel values in the sub-scanning direction is counted from the first line (S6 to S12). Specifically, DAT (0,0) and DAT (1,0) ,. ． ． , DAT (0,3) and DAT
(1,3) are sequentially compared, and then DAT (1,0) and DAT (2,3) are set between the second and third lines.
0) ,. ． ． , DAT (1,3)) and DAT (2,3)
Of the DAT (2,0) and DAT (3,0) ,. ． ． , D
The AT (2,3) and the DAT (3,3) are compared in order, and the counter COUTA is incremented each time the comparison is disagreement.

Next, the counters I and J are cleared to zero (S1
After 3), for the first column (I = 0), the values of the pixels adjacent in the sub-scanning direction are compared, and the disagreement points are counted by the counter COUTB (S14 to S17). Following this, from the first column in the main scanning direction, DAT
(0,0) and DAT (1,0) ,. ． ． , DAT (2
0) and DAT (3,0), DAT (0,0) and DAT
(0,1) ,. ． ． , DAT (3,0) and DAT (3,
1), DAT (0,1) and DAT (0,2) ,. ． ． ,
DAT (3,1) and DAT (3,2), DAT (0,
2) and DAT (0,3) ,. ． ． , DAT (3,2) and DAT (3,3) are compared in order, and the counter COUNTB is incremented for each mismatch (S19 to S24).

Then, the counters COUNTA, COUNT
The TB values are both predetermined reference values S (S = 4 in this embodiment).
When the above is the case (when the determination results of S25 and S26 are both YES), it is determined that the current pixel of interest is "no correlation" (the number of pixel value change points is small), and "
1 "signal is output (S27). Otherwise, a
= “0”.

The area determination unit 3 counts up in synchronization with the clock RCLKI if the non-correlation signal a is "1" (no correlation), but clears it if a = "0" (correlation). . That is, the number of consecutive "non-correlated" pixels is counted. Then, the count value is a predetermined reference value S
When (S = 4 in this embodiment) or more, the discrimination signal b = "1"
Output a signal (judged to be a photographic image), b otherwise
= “0” is output (determined as a character image).

The above image discriminating operation will be described with reference to FIGS. In the example of the character image shown in (a) of FIG. 5 (this is the same as the image shown in (a) of FIG. 10), the image discrimination of the pixel of the star is considered. Three pixels before the pixel of interest, the reference area is at the position shown in (b), so the value of COUTNA is 5 and the value of COUNTB is 2 in the correlation calculation, and the result of the correlation calculation is “correlation”, a = "
0 "is output, and the area determination unit (counter) 3 is cleared here. From the next pixel to the pixel of interest, the reference areas are at positions as shown in (c), (d), and (e), respectively. And the value of COUNTA and COUNTB are both 4
Because of the above, the result of the correlation calculation is “no correlation”, and a = “1” is output. Therefore, the count value of the area determination unit 3 at the time of the pixel of interest becomes 3, the image is determined to be a character image, and b = “0” is output.

Similarly, in the example of the photographic image shown in FIG. 6A (this is the same as the image shown in FIG. 10B), the image discrimination of the pixel of the star mark is considered. (B) to (e)
As shown in, all the correlation calculation results for the four pixels from three pixels before to the target pixel are “no correlation”,
Since a = “1” is continuously output, the count value of the area determination unit 3 increases to 4 at the time of the pixel of interest and is determined to be a “photographic image”, so b = “1” is output. It

The character image of FIG. 5 (a) and the photographic image of FIG. 6 (b) cannot be identified by the single reference of the 4 × 4 matrix as explained with reference to FIG. Although it was necessary to expand the reference area to 4 × 8, according to the present invention, the identification can be performed with the 4 × 4 reference area. This is because, in the present invention, the reference calculation area is enlarged because the correlation calculation results for a plurality of pixels before that are also used for the determination of the target pixel.

When the output of the image data for one page from the storage unit 1 is completed, the control unit 6 stops the clock RCLK and outputs the recording start signal PSTR. Recording unit 5
When the recording start signal PSTR is input, the clock PCLK is output to sequentially read and record the correction data from the storage unit 4.

According to the second embodiment of the present invention, the area determining section 3 of FIG. 1 has the configuration shown in FIG. In FIG.
Reference numerals 21 and 22 are counters, and 23 is a latch circuit. a,
b, RCLK, and CLR are signals corresponding to the signals of the same name shown in FIG.

One counter 21 has a clock RCLK.
In synchronization with, the counter is incremented when the non-correlation signal a is "1" (no correlation), and cleared when a = "0" (correlation is present). That is, the counter 21 counts the number of consecutive "non-correlated" pixels. Then, when the count value exceeds a predetermined reference value S (S = 4 in this embodiment), the output signal c is set to "1".

The other counter 22 has a clock RCLK.
In synchronization with, the counter is incremented when the non-correlation signal a is "0" (correlation is present), and cleared when a = "1".
That is, the number of consecutive "correlated" pixels is counted.
The count value is the reference value S (S = 4 in this embodiment).
When the above is reached, the output signal d is set to "1".

The latch circuit 23 has c = “1” and d = ”.
When 0 ", the discrimination signal b is set to" 1 "(photo image), and c
When == "0" and d = "1", the discrimination signal b is set to "0" (character image). Both signals c and d are "0" or "
When it is 1 ", the state of the discrimination signal b does not change.

According to the configuration of the area determining unit 3 of the first embodiment, when even one pixel "correlated" appears in the area where "no correlation" pixels continue, the determination signal b is generated at that position. Becomes "0". On the other hand, according to the configuration of the area determination unit 3 of the present embodiment, there are 4 “no correlation” pixels.
When four or more pixels with "correlation" are consecutive after one or more consecutive pixels, or when four or more pixels with "no correlation" are consecutive after four or more pixels with "correlation",
The state of the discrimination signal b changes, and in the other cases, the previous state is held, so that more stable photo / character discrimination is possible.

According to the third embodiment of the present invention, the area determination section 3 of FIG. 1 has the configuration shown in FIG. In FIG.
Reference numeral 31 is a register in which (L-1) latch circuits are connected in series, 32 is an adder, and 33 is a comparator. In this embodiment, L is equal to the total number of pixels in one line. a, b, RC
LK and CLR are signals corresponding to the signals of the same name in FIG.

The decorrelation signal a is input to the register 31,
It is sequentially shifted to the latch circuit in the subsequent stage in synchronization with the clock RCLK. The input decorrelation signal a and the output signal (delayed signal a) of the latch circuit of each stage of the register 31 are input to the adder 32, and the number of "1" bits in the input L bit signal is added. It is output from the device 32. This added value (the number of pixels having "no correlation") is input to the comparator 33 and compared with a predetermined reference value Q, and when the added value is equal to or greater than the reference value Q, the discrimination signal b becomes "1". However, otherwise b = “0”.

As described above, in this embodiment, it is not the number of consecutive "correlated" pixels, but the number of "correlated" pixels included in the L consecutive pixels that is greater than the reference number. Character judgment is performed. With this method, it is easy to avoid the influence of noise, and it is possible to perform stable image discrimination in large units (for example, in units of one line).

However, in the present embodiment, the number of pixels (pixels of a = "0") determined to have few pixel value changes is counted, and similar image discrimination is performed based on this count value. It is also possible to do so.

In the dither image, a pixel having a low or high black ratio in the reference area may be determined as a character image area (however, such a pattern hardly appears in the character image). In order to perform more accurate image determination, in addition to the above correlation calculation, the correction table unit 2 determines the black ratio of the reference area, or determines the specific pattern of the photographic image or the character image. And is effective. Further, the image identifying apparatus according to the present invention can be applied to various purposes other than the image recording apparatus.

[0049]

As is apparent from the above description, the present invention has the following effects. (1) In the first to third inventions, it can be considered that the substantial size of the reference area is larger than that of the reference area of each pixel. Therefore, the size of the reference area is minimized. Thus, it is possible to discriminate an image that could not be discriminated by the size of the reference area. (2) In the photographic image area, pixels with few pixel value change points may locally appear during continuous pixels with many pixel value change points. A second aspect or a third aspect of the invention is an area of pixels having a small number of locally changed pixel values in a photographic image area, or a pixel having a large number of locally changed pixel values in a character image area. It is possible to correctly discriminate the area of. (3) In the third aspect of the invention, stable image discrimination can be performed while avoiding the influence of noise, and image discrimination can be stably performed in large units.

[Brief description of drawings]

FIG. 1 is a block diagram showing a first embodiment of the present invention.

FIG. 2 is an explanatory diagram of reference pixels and correction data.

[Fig. 3] Operation timing diagram

FIG. 4 is a flowchart of correlation calculation.

FIG. 5A is a diagram showing an example of a character image, and FIGS. 5B to 5E are diagrams showing a process of discriminating character images.

FIG. 6A is a diagram showing an example of a photographic image. FIGS. 6B to 6E are diagrams showing a process of discriminating photographic images.

FIG. 7 is a block diagram of an area determination unit according to a second embodiment of the present invention.

FIG. 8 is a block diagram of an area determination unit according to a second embodiment of the present invention.

FIG. 9 is a diagram showing an example of a pseudo halftone expression by the dither method in (a) to (q).

10A is a diagram showing an example of a character image. FIG. 10B is a diagram showing an example of a photographic image. FIG. 10C is a diagram showing a pattern of a 4 × 4 reference area of a character image. (E) A diagram showing a 4 × 8 reference area pattern of a character image (f) A diagram showing a 4 × 8 reference area pattern of a photographic image

[Explanation of symbols]

 1, 4 storage unit 2 correction table unit 3 area determination unit 5 recording unit 6 control unit 8 image data input terminal 9 clock output terminal a non-correlation signal b determination signal 21,22 counter 23 latch circuit 31 register 32 adder 33 Comparator

Claims (3)

[Claims] 1. A means for deciding whether or not there are a large number of changing points of a pixel value in a reference area with respect to a reference for each pixel of binary image data, and a large number of changing points by this means. An image discriminating apparatus having means for counting the number of consecutively appearing pixels, deciding that it is a photographic image when the counted value is a predetermined value or more, and deciding that it is a character image otherwise. 2. For each pixel of the binary image data, there is a judging means for judging whether there are a large number or a small number of changing points of the pixel value in the reference area with respect to a reference, and if there are many changing points by this judging means. First counting means for counting the number of consecutively appearing pixels that have been determined, and second counting means for counting the number of consecutively appearing pixels that have been determined by the determining means to have few changes. When the count value of the first counting means is equal to or more than a predetermined value and the count value of the second counting means is less than a predetermined value, it is determined to be a photographic image, and the count value of the first counting means is a predetermined value. An image discrimination device having means for determining that the character image is less than the second count value is equal to or more than a predetermined value and determining the same as the immediately preceding pixel under other conditions. 3. A judging means for judging whether each pixel of the binary image data has a pixel value change position in the reference area is larger or smaller than a certain reference, and among a predetermined number of consecutive pixels. An image discriminating apparatus having a unit for discriminating whether the image is a photographic image or a character image based on the number of pixels determined by the determining unit to have a large number or a small number of changed pixel values.