JP3474431B2 - Image processing method - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method for discriminating the image type of each pixel included in an image to be image processed.

[0002]

2. Description of the Related Art For example, in a facsimile machine which transmits an image signal via a public telephone line network, a non-halftone image such as a character is converted into a non-halftone binary image data on a transmitting side by using a predetermined threshold value. The converted halftone image such as a photograph is converted into binary halftone image data of pseudo halftone by a dither method or the like and is transmitted to the reception side. On the other hand, the receiving side needs to perform different image processing depending on the non-halftone binary image data and the pseudo halftone image data. Therefore, the receiving side automatically determines whether each pixel of the received binary image data is non-halftone binary image data or pseudo halftone image data. .

As a conventional method for discriminating the image type of each pixel included in the image to be image-processed as described above, according to the inventions disclosed in JP-A-4-361475 and JP-A-5-91312, An area of a predetermined range including the pixel of interest is set in the image to be processed, and the image type of the pixel of interest is determined based on the adjacency state of pixels of the same type in this area. Specifically, by counting the number of adjacencies with peripheral pixels of the same type for each pixel included in a region of a predetermined range set in the image, and comparing the total number of adjacencies in this region with a predetermined threshold value. The image type of the pixel of interest is selected.

[0004]

However, in any of the above-mentioned conventional methods for discriminating the image type of each pixel included in the image to be processed, the target pixel set in the image to be processed includes the target pixel. The image type of the pixel of interest is determined by counting the number of adjacent pixels for all pixels belonging to the area of the predetermined range and comparing the total number of adjacent pixels with a predetermined threshold to determine the image type of the pixel of interest. There is a problem that the process for making the determination is complicated and requires a long time. Such a problem is not limited to the case of determining the image type of a pixel included in a binarized image, but the image of a pixel included in multi-tone image data in which the density of each pixel is represented by, for example, 8-bit density data. The same occurs when determining the type.

The object of the present invention is to determine the image type of the pixel of interest by comparing the connection probability of pixels of the same type within a predetermined range of area including the pixel of interest set in the image with a threshold value. An image processing method capable of determining the image type of a pixel of interest without counting the number of adjacent pixels of one type for each pixel and simplifying and shortening the process of determining the image type of the pixel of interest To provide.

[0006]

According to a first aspect of the present invention, a first area composed of a constant number of pixels centering on a target pixel which is an image type discrimination target is used as an input image. Within the first area and the image forming the first area within this first area.
A plurality of first pixels composed of a smaller number of pixels than the prime number.
2 areas are set, and the same in each of the second areas
The number of types of pixels is counted as the pixel pattern, and this
Concatenation for each of the second regions corresponding to the numerical result
Sum of probabilities for each number of pixels of the same type in the first region
It is characterized in that the image type of the pixel of interest is discriminated by comparing with the threshold value set in advance .

In the invention described in claim 1, attention is paid to
A plurality of second pixels set in the first area centered on the pixel
A series corresponding to the same number of pixels in each area
By comparing the sum of the connection probabilities with a threshold,
The image type is determined. Therefore, focusing on the pixel of interest
Based on multiple pixel patterns in the first region
The image area of each pixel of interest is determined from the calculated connection probability.
As a result, the accuracy of discriminating the image area is improved.

According to a second aspect of the present invention, with respect to an input image having multi-tone density information for each pixel, the multi-tone density information is divided into a plurality of ranges, and all or a part of the plurality of ranges is divided. the second region about the pixels included in each Noso
Put the sum of the connection probabilities for each in the first region
Compare with a predetermined threshold for each number of pixels of the same type
Thus, the image type of the pixel of interest is determined.

[0009] In the invention as set forth in claim 2, each pixel of the input image if it has a density information of multi-gradation, the pixels included in each of all or some of the plurality of concentration ranges
The image type of the pixel of interest is determined based on the sum of the connection probabilities corresponding to the number of pixels . Therefore, even when the input image is a multi-valued image, the duplication in the first region including the pixel of interest is performed.
The image type of the pixel of interest is determined from several pixel patterns.

[0010] invention as set forth in claim 3, the input image having a density binary information for each pixel, the second <br/> about the pixels having either one of the density information of the two values The sum of the connection probabilities for each of the regions is calculated as the first region.
With a threshold value that is predetermined for each number of pixels of the same type within
The feature is that the image type of the pixel of interest is determined by comparing .

According to the third aspect of the invention, when each pixel of the input image has binary density information, for example, the number of black pixels in the area including the target pixel is
The image type of the pixel of interest is determined based on the corresponding connection probability . Therefore, even when the input image is a binary image, the image type of the target pixel is determined from the plurality of pixel patterns in the first region including the target pixel.

[0012]

[0013]

According to a fourth aspect of the present invention, the number of pixels in the first area is calculated based on the total number of pixels counted in the second area, and the result of this calculation is compared with the total sum of connection probabilities. It is characterized by determining a threshold value.

According to the invention described in claim 4 ,
The threshold value used to determine the image type of the target pixel is determined from the number of pixels in the first area calculated based on the result of counting the number of pixels in the area. Therefore, it is not necessary to count the number of pixels in the first region when determining the threshold value,
The result of counting the number of pixels in the second area is effectively used.

The invention described in claim 5 is characterized in that each of the second regions is set without overlapping pixels in the first region.

In the invention described in claim 5 , the first
The image type of the pixel of interest is determined from the connection probability calculated based on the respective pixel patterns of the plurality of second regions set without overlapping pixels in the region.
Therefore, the second number of regions is efficiently set in the input image, and the region referred to in determining the image type is expanded, and the determination accuracy is further improved.

According to a sixth aspect of the present invention, when the magnification conversion processing is performed on the input image, the information plane constituted by the connection probability for each pixel of interest is subjected to the magnification conversion processing with the same magnification as the input image. It is characterized in that the image type of the pixel of interest is discriminated based on the connection probability of each pixel forming the information plane after the magnification conversion process, and the processing according to the discrimination result is executed.

According to the sixth aspect of the present invention, when the magnification conversion processing is performed on the input image, the information plane constituted by the connection probability calculated for each pixel of the input image is the magnification of the input image. The conversion is performed with the same scaling ratio as that in the conversion process. Therefore, during the magnification conversion processing of the input image in the resolution conversion or the like accompanying the adjustment of the output image size, the concatenation of each pixel forming the information plane after the conversion with the same scaling ratio as that in the magnification conversion processing of the input image is performed. The image type of each pixel after the magnification conversion process is accurately determined using the probability.

According to a seventh aspect of the present invention, at the time of the magnification conversion process for the input image, the information plane constituted by the image type discriminated for each pixel of interest based on the connection probability is the same as the scaling ratio of the input image. It is characterized in that the magnification conversion processing is performed with a magnification change ratio, and the processing according to the image type for each pixel forming the information plane after the magnification conversion processing is executed.

In the invention described in claim 7 , when the magnification conversion processing is performed on the input image, the information plane formed by the image type previously determined based on the connection probability for each pixel of the input image is The input image is converted with the same scaling factor as the scaling factor in the scaling process. Therefore, during the magnification conversion processing of the input image in the resolution conversion and the like accompanying the adjustment of the output image size, the magnification conversion of the information plane configured by the value specifying the image type having the extremely small amount of data compared to the connection probability of each pixel is performed. Since the processing is performed, the burden of data storage and calculation is reduced as compared with the case where the magnification conversion processing is performed on the information plane configured by the connection probability for each pixel.

[0022]

BEST MODE FOR CARRYING OUT THE INVENTION A method for discriminating the image type of each pixel having 8-bit density information in an image to be image-processed will be described below.

FIG. 1 is a flow chart showing the processing procedure of the image processing method according to the embodiment of the present invention. In the image processing method according to the embodiment of the present invention, the target pixel A to be image-processed in the input image is specified (s1),
A large matrix M of 3 × 3 pixels (corresponding to the first region of the present invention) centered on the specified target pixel A is set (s2), and a plurality of small matrices included in the large matrix M are set. m (corresponding to the second area of the present invention) is set (s3).

The small matrix m is, for example, a matrix of 2 × 2 pixels including the target pixel A. When the small matrix m is 2 × 2 pixels, four small matrices m are included in the large matrix M. Is set. In each of the small matrices m, the number of pixels of the same type having specific density information is counted (s4), the connection probability Z of pixels corresponding to the counted number of pixels is calculated (s5), and the calculated connection probability Z is calculated. Is compared with a threshold value to specify the type of image area of the target pixel A (s6, s7).

FIG. 2 is a diagram for explaining an image processing method according to the first embodiment of the present invention. In the image processing method according to this embodiment, when the pixel located at the coordinates (x, y) in the image G to be image-processed is identified as the target pixel A in s1 shown in FIG. 1, s2 shown in FIG. A large matrix M of 3 × 3 pixels centering on the pixel of interest A is set in the processing of 3), and four small matrices ma to md of 2 × 2 pixels are set in the large matrix M in the processing of s3.
Is set (see FIG. 2A).

The small matrix md is the coordinates (x, y), (x + 1, y), (x, y in the image G to be image-processed.
It includes four pixels located at (y + 1) and (x + 1, y + 1). As shown in FIG. 2B, this small matrix m is set one pixel at a time in the x direction from the pixel at coordinates (0, 0), and when reaching the end in the x direction of the image, one pixel is formed in the y direction. The pixel is moved and sequentially set one pixel at a time from the coordinate (0, 1) in the x direction.

The counting of the number of pixels of the same type having the specific density information in s4 shown in FIG. 1 is performed, for example, with respect to the density information of 8 bits (0 to 255) which is the input level of the image G, the density class C0 (0 -J) Density class C1 (j + 1 to k) Density class C2 (k + 1 to 255) (however, 0 <j <k <255) three-level density classes are set, and density classes C0 to C0 for each of the small matrices m. The number of pixels S0 to S2 included in each of C2 is counted.

That is, the coordinates (x, y) shown in FIG.
The target pixel A of 3 × 3 centered on the target pixel A
4 small matrices ma belonging to the large matrix M of
˜md, the number of pixels belonging to the density classes C0 to C2 in each of the small matrices ma to md is calculated,
These are S0ma, S0mb, S0mc, S0md S1ma, S1mb, S1mc, S1md S2ma, S2mb, S2mc and S2md. Considering that each of the small matrices ma to md is composed of 2 × 2 pixels, each of the number of images Snma to Snmd (n = 0 to 2) is 0.
Takes a value of ~ 4.

The number of pixels thus obtained Snma
Using Snmd, the connection probability Z, which is the probability that each pixel belonging to the large matrix M is adjacent to the target pixel A, is defined as the sum of the connection probabilities z for each number of pixels of the small matrix m included in the large matrix M, and the density class It is calculated for each of C0 to C2. That is, the connection probability z of the small matrix level is determined in advance according to the number of pixels in the small matrix m, and for each of the four small matrices ma to md included in the large matrix M centered on the pixel of interest A. Based on the counted number of pixels, the small matrix ma ~
The connection probabilities za to zd of md are determined, and the connection probability Z at the large matrix level is calculated as the sum of these. Since the pixel connection state may be connected not only in the vertical and horizontal directions but also in the oblique direction, the connection probability Z is the connection probability Z in only the vertical and horizontal directions and the connection probability Z ′ including the diagonal direction. calculate. Therefore, the connection probability z only in the vertical and horizontal directions and the connection probability z'including the diagonal direction are determined even at the small matrix level.

A method of determining the connection probabilities z and z'at the small matrix level will be described below. The connection probabilities z and z ′ at the small matrix level for each of the density classes C0 to C2 for the specified pixel of interest A are the small matrix m.
In the case where the number of connected pixels included in a specific density class in the pixel is Ta, the number of peripheral pixels outside the small matrix m is p, and all peripheral pixels outside the small matrix m are in the same density class, the small matrix m Let Tb be the number of connected specific pixels and peripheral pixels in the small matrix ma
For each of ˜md, z = Ta + (Tb / p) ... Equation 1 z '= Ta' + (Tb '/ p') ... Equation 2

However, when the number of pixels is 1 to 3, a plurality of patterns may be considered for each number of pixels of a specific density class in the small matrix m. In this case, the first term on the right side of the above equation 1 and As the second term, an average of a plurality of patterns is calculated, and z = ΣTa / q + Σ (Tb / p) / q Equation 3 z ′ = ΣTa ′ / q + Σ (Tb ′ / p ′) / q Equation 4 (However, q is the number of patterns in the small matrix).

As a result, for example, in the small matrix md for the target pixel A, the connection probabilities z and z'at the small matrix level are determined as shown in FIG.
That is, a pattern as shown in FIG. 3 is considered for each of the cases where the number of pixels Snmd is 0 to 4, and the number of pixels Snmd is
When d = 0, both the connection probabilities z and z ′ are naturally 0. When the number of pixels Snmd = 1, in each of the four types of patterns a to d shown in FIG. 3, the number of connected pixels Ta in the small matrix is 0, and the number of connected pixels in the vertical and horizontal directions with the peripheral pixels. Tb is 0, 1, 2, 1 in order from the pattern a, and the number of connections T including the diagonal direction T
b'is 0, 2, 5, and 2 in order from the pattern a. In each of these four types of patterns a to d, the number of peripheral pixels p in the vertical and horizontal directions is 4, and the number of peripheral pixels p ′ including the diagonal direction is 5. Substituting these values into Equations 3 and 4 above, Is obtained.

Further, when the number of pixels Snmd = 2, in each of the six types of patterns a to f shown in FIG.
The number of connections Ta in the vertical and horizontal directions in the small matrix md is 1, 1, 0, 0, 1, 1 in order from the pattern a, and the number of connections T including the diagonal direction in the small matrix md is T.
a ′ is 1 for all of the patterns a to f, and the number of connections Tb in the vertical and horizontal directions with peripheral pixels is 1 in order from the pattern a,
1, 2, 2, 3, 3 and the number of connections T including the diagonal direction T
b'is 2, 2, 5, 4, 5, 5 in order from the pattern a. In each of these six types of patterns a to f, the number of peripheral pixels p in the vertical and horizontal directions is 4, and the number of peripheral pixels p ′ including the diagonal direction is 5. Substituting these values into the above equations 3 and 4, z = (1 + 1 + 0 + 0 + 1 + 1) / 6 + (1/4 + 1/4 + 2/4 + 2/4 + 3/4/4/3) / 6 = 7/6 z '= (1 × 6 ) / 6 + (2/5 + 2/5 + 5/5 + 4/5 + 5/5 + 5/5) / 6 = 53/30.

Further, when the number of pixels Snmd = 3,
In each of the four types of patterns a to d shown in FIG. 3, the number of connections T in the vertical and horizontal directions in the small matrix md
a is 2 in all, the number of connections Ta ′ including the diagonal direction in the small matrix md is 3, and the numbers of connections Tb in the vertical and horizontal directions with peripheral pixels are 2, 3, 3, 4 in order from the pattern a. And the number of connections Tb ′ including the diagonal direction is 4, 5, 5, 5 in order from the pattern a. In each of these four types of patterns a to d, the number of peripheral pixels p in the vertical and horizontal directions is 4, and the number of peripheral pixels p ′ including the diagonal direction is p ′.
Is 5. Substituting these values into Equations 3 and 4 above, Is obtained.

In addition, when the number of pixels Snmd = 4,
The number Ta of connections in the vertical and horizontal directions in the small matrix md is 4, the number of connections Ta′6 including the diagonal direction in the small matrix md, and the number of connections Tb in the vertical and horizontal directions with the peripheral pixels is 4 and the diagonal direction. The number of connections Tb ′ including 5 is 5. In this case, the number of peripheral pixels p in the vertical and horizontal directions is 4,
The peripheral pixel number p ′ including the diagonal direction is 5. Substituting these values into Equation 1 and Equation 2 above, Is obtained.

The small matrix level connection probabilities z and z'determined as described above are stored in advance for each number of pixels in the small matrix and are included in the large matrix M in the image to be image-processed. The connection probabilities z and z'corresponding to the pixel numbers Snma to Snmd of the respective small matrices ma to md are read out, and the sums of the read connection probabilities z and z'are respectively connected to the large matrix level connection probabilities Z and Z '. And

The connection probabilities Z and Z'are given to the concentration class C0.
To C2, and the image type of the target pixel A located at the center of the large matrix M is determined based on the calculated result. For example, when the number of pixels in the set large matrix M is 4 for any of the density classes, the pattern shown in FIG. Becomes On the other hand, in the pattern shown in FIG. 4B, Z = 11/4 + 1/4 + 11/4 + 7 / 6≈6.9 Z ′ = 59/20 + 9/20 + 59/20 + 53 / 30≈8.1.

Therefore, even if the number of pixels of a particular density level belonging to the large matrix M centered on the pixel of interest A is the same, the connection probabilities Z and Z'are different depending on the pixel pattern. Therefore, the calculated connection probabilities Z and Z '
By comparing with the threshold value, the image type of the target pixel A can be determined. For example, when the number of pixels in the large matrix M is 4 for the density class C2,
The threshold value T for comparison with the connection probability Z is set to 6, and the threshold value T'for comparison with the connection probability Z'is set to 7, and Z ≧ T and Z ′.
If ≧ T ′, it is determined that the image type of the target pixel A located in the center of the large matrix M is a character image, and if Z <T and Z ′ <T ′, the center of the large matrix M is focused. It can be determined that the image type of the pixel A is a halftone image.

The image type of the target pixel A may be determined by performing the calculation process and the comparison process with the threshold value for only one of the connection probability Z and the connection probability Z'in the large matrix M.

Since the values of the threshold values T and T'compared with the connection probabilities Z and Z'are set according to the number of pixels belonging to a specific density class in the large matrix M, the image of the target pixel A is displayed. Large matrix M for type discrimination
It is necessary to know the number of pixels belonging to the specific density class in. At this time, by calculating the number of pixels in the large matrix M from the counting result of the number of pixels in the four small matrices ma to md set in the large matrix M, only the information obtained from the small matrices ma to md is used. The image type of the target pixel A can be determined. In this case, four small matrices ma
It is conceivable that the total sum of the numbers of pixels in the four small matrices ma to md is corrected by a predetermined correction value in consideration of the fact that there are pixels that are counted in duplicate in up to md.

Furthermore, the range of each small matrix m can be determined when the image to be image-processed is input in the state shown in FIG. 2 (B). Therefore, when an image to be image-processed is input, the number of pixels in each small matrix m is counted and stored in advance, and the stored content is appropriately read out for each target pixel A whose image type is to be determined, and the connection probability Z and It may be used to calculate Z ′.

FIG. 5 is a diagram for explaining an image processing method according to the second embodiment of the present invention. In the image processing method according to this embodiment, when the pixel located at the coordinates (x, y) in the image G to be image-processed is identified as the target pixel A in s1 shown in FIG. 1, s2 shown in FIG. In the processing of 3), a large matrix M of 3 × 3 pixels centering on the pixel of interest A is set, and in the processing of s3, three small matrices me to mg of 3 × 1 pixels are included in the large matrix M.
Is set (see FIG. 5A).

For example, the small matrix mf is the coordinates (x-1, y), (x, y) in the image G to be image-processed.
And three pixels located at (x + 1, y). As shown in FIG. 5B, the small matrices me to mg are set by moving 3 pixels in the x direction from the pixel at the coordinates (0, 0), and when reaching the end portion in the x direction in the image, they are set in the y direction. 1
The pixel is moved to move from the coordinate (0, 1) by 3 pixels in the x direction and set.

The count of the number of pixels having the specific density information in s4 shown in FIG. 1 is, for example, 8 bits (0 to 0) which is the input level of the image G, as in the first embodiment.
255), three-level density classes C0 to C2 are set for the density information, and the small matrix me to
The number of pixels S0 to S2 included in each of the density classes C0 to C2 is counted for each mg.

That is, the coordinates (x, y) shown in FIG.
The target pixel A of 3 × 3 centered on the target pixel A
3 small matrices me belonging to the large matrix M of
~ Mg is selected, the number of pixels belonging to the density classes C0 to C2 in each of the small matrices me to mg is calculated,
These are S0me, S0mf, S0mg S1me, S1mf, S1mg S2me, S2mf, S2mg. Considering that each of the small matrices me to mg is composed of 2 × 2 pixels, each of the number of images Snme to Snmg (n = 0 to 2) is 0.
Takes a value of ~ 3.

The number of pixels Snme obtained in this way
Using Snmg, the connection probability Z, which is the probability that each pixel belonging to the large matrix M is adjacent to the target pixel A, is taken as the sum of the connection probabilities z for each number of pixels of the small matrix m included in the large matrix M, and the density class It is calculated for each of C0 to C2. That is, as in the first embodiment, the connection probability z of the small matrix level is determined in advance according to the number of pixels in the small matrix m, and is included in the large matrix M centered on the target pixel A. Based on the number of pixels counted for each of the small matrices me to mg, the connection probability z of each of the small matrices me to mg
e to zg and ze 'to zg' are determined, and the sum of these is calculated as the connection probabilities Z and Z'at the large matrix level.

In the image processing method according to this embodiment, the connection probabilities z and z'at the small matrix level are expressed by the above equation (3).
And Equation 4 are used to determine as shown in FIG. That is, a pattern as shown in FIG. 6 can be considered for each of the cases where the number of pixels is 0 to 3, and when the number of pixels is 0,
Naturally, the connection probabilities z and z'are both 0. When the number of pixels is 1, in each of the nine types of patterns a to i shown in FIG. 6, the number of connected pixels Ta in the small matrix is 0, and the number of connected pixels Tb in the vertical and horizontal directions with the peripheral pixels. Is 1, 1, 1, 2, 2, in order from the pattern a.
2, 1, 1, 1, and the number of connections Tb ′ including the diagonal direction is 2, 3, 2, 4, 6, 4, 2, 3, in order from the pattern a.
It is 2. In each of these nine types of patterns a to i, the number of peripheral pixels in the vertical and horizontal directions, p, and the number of peripheral pixels in the oblique direction, p ′, are all 6. Substituting these values into the above equations 3 and 4, z = 0 + (1/6 × 6 + 2/6 × 3) / 9 = 2/9 z ′ = 0 + (2/6 × 4 + 3/6 × 2 + 4/6 X2 + 6/6) / 9 = 14/27 is obtained.

Further, when the number of pixels is 2, in each of the nine types of patterns a to i shown in FIG. 6, the number Ta of connections in the vertical and horizontal directions and the number Ta 'of connections including the oblique direction in the small matrix are both. Also from pattern a in order 1,
0, 1, 1, 0, 1, 1, 0, 1 and the number of connection Tb in the vertical and horizontal directions with peripheral pixels is 2, 2, in order from the pattern a.
2, 4, 4, 4, 2, 2, 2, and the number of connections Tb ′ including the diagonal direction is 3, 3, 3, 6, 6, in order from the pattern a.
6, 3, 3, and 3. These nine types of patterns a to i
In each of the above, the number of peripheral pixels p in the vertical and horizontal directions, and
The number of peripheral pixels p ′ including the diagonal direction is 6 in all cases. Substituting these values into the above equations 3 and 4, z = (1 × 6) / 9 + (2/6 × 6 + 4/6 × 3) / 9 = 10/9 z ′ = (1 × 6) / 9 + (3/6 × 6 + 6/6 × 3) / 9 = 4/3 is obtained.

Further, when the number of pixels is three, in each of the three types of patterns a to c shown in FIG. 6, the number Ta of connections in the vertical and horizontal directions and the number Ta 'of connections including the diagonal direction in the small matrix are any. Is also 2, and the number of connections Tb in the vertical and horizontal directions with the peripheral pixels and the number of connections Tb ′ including the oblique direction.
Are 3, 6, and 3 in order from the pattern a. The number p of peripheral pixels in the vertical and horizontal directions and the number p'of peripheral pixels including the oblique direction in each of these three types of patterns a to c are 6. Substituting these values into Equations 3 and 4 above, Is obtained.

The connection probabilities z and z'at the small matrix level determined as described above are stored in advance for each number of pixels in the small matrix and are included in the large matrix M in the image to be image-processed. The connection probabilities z and z'corresponding to the pixel numbers Snme to Snmg of the respective small matrices me to mg are read out, and the respective sums of the read connection probabilities z and z'are connected to the large matrix level connection probabilities Z and Z '. And

For example, when the number of pixels in the set large matrix M is 4, the pattern shown in FIG. Becomes On the other hand, in the pattern shown in FIG. Becomes

Therefore, even if the number of pixels of a specific density level belonging to the large matrix M centering on the pixel of interest A is the same, the connection probabilities Z and Z'are different depending on the pixel pattern. Therefore, the calculated connection probabilities Z and Z '
It is possible to determine the type of the image area to which the target pixel A belongs by comparing with the threshold value. For example, when the number of pixels in the large matrix M is 4, the threshold value T to be compared with the connection probability Z is set to 2, and if Z ≧ T, the target pixel A located at the center of the large matrix M is in the character area. It is determined that the pixel is included, and if Z <T, the large matrix M
It can be determined that the pixel of interest A located at the center of is the pixel included in the halftone area.

Further, in the image processing method according to this embodiment, 3 in the large matrix M of 3 × 3 pixels.
Since the small matrices me to mg are set without overlapping the pixels, the same pixel is not counted a plurality of times when the number of pixels belonging to the three small matrices me to mg is counted. Therefore, the sum of the pixel count values in the three small matrices me to mg can be directly used as the pixel count values in the large matrix M, and the connection probability calculation process can be simplified.

Furthermore, since each of the plurality of small matrices m is set in the large matrix M without overlapping pixels, the image range used for determining the image type of the target pixel A is expanded, and as a result, the image type The discrimination accuracy is improved.

FIG. 8 is a diagram for explaining an image processing method according to the third embodiment of the present invention. In the image processing method according to this embodiment, when the pixel located at the coordinates (x, y) in the image G to be image-processed is identified as the target pixel A in s1 shown in FIG. 1, s2 shown in FIG. In the processing of, from coordinates (x-3, y-3) to (x + 4, y +
A large matrix M of 8 × 8 pixels centering on the target pixel A is set in the range of 4), and 16 small matrices m00 of 2 × 2 pixels in the large matrix M in the processing of s3.
To m33 are set so that pixels do not overlap in each small matrix.

The counting of the number of pixels having the specific density information in s4 shown in FIG. 1 is, for example, 8 bits (0 to 0) which is the input level of the image G, as in the first embodiment.
255) density information, density class C0 (0 to j) density class C1 (j + 1 to k) density class C2 (k + 1 to 255) (where 0 <j <k <255) density levels are set. And small matrix m0
The number of pixels S0 to S2 included in each of the density classes C0 to C2 is counted for each of 0 to m33.

That is, the coordinates (x, y) shown in FIG.
8 × 8 centered on the target pixel A
16 small matrices m belonging to the large matrix M of
00-m33 is selected and the small matrix m00-m33 is selected.
S0m00, ..., S0m03, S0m10, ..., S0m33 S1m00, ..., S1m03, S1m10, ..., S1m33 S2m00, ..., S2m03, S2m10, ..., S2m33. Considering that each of the small matrices m00 to m33 is composed of 2 × 2 pixels, each of the image numbers Snm00 to Snm33 (n = 0 to 2) takes a value of 0 to 4.

The number of pixels Snm00 thus obtained
~ Snm33 is used, the connection probability Z, which is the probability that each pixel belonging to the large matrix M is adjacent to the target pixel A, is defined as a small matrix m00 to m3 included in the large matrix M.
The density class C is the sum of the connection probabilities z for each number of 3 pixels.
It is calculated every 0 to C2. That is, according to the number of pixels in the small matrix m, the connection probability z at the small matrix level
The advance decision, 16 small matrices m00~m contained in a large matrix M centering on the target pixel A
Based on the number of pixels counted for each of 33, the connection probabilities z and z ′ of the small matrices m00 to m33 are determined, and the connection probabilities Z and Z ′ at the large matrix level are calculated as the sum of these.

In the image processing method according to this embodiment, the connection probabilities z and z'at the small matrix level are expressed by the above equation (3).
And Equation 4 are used to determine as shown in FIG. That is, a pattern as shown in FIG. 9 can be considered for each of the cases where the number of pixels is 0 to 4, and when the number of pixels is 0,
Naturally, the connection probabilities z and z'are both 0. When the number of pixels is 1, in each of the four types of patterns a to d shown in FIG. 9, the number of connected pixels Ta in the small matrix is 0, and the number of connected pixels Tb in the vertical and horizontal directions with peripheral pixels. Is 2, and the number Tb 'of connections including the diagonal direction is 5. These four types of patterns a
In each of the patterns d to d, the number of peripheral pixels p in the vertical and horizontal directions is 8, and the number of peripheral pixels p ′ including the diagonal direction is 12. Substituting these values into Equations 3 and 4 above, Is obtained.

Further, when the number of pixels is 2, in each of the six types of patterns a to f shown in FIG. 9, the number Ta of connections in the vertical and horizontal directions in the small matrix is 1, 1, 0, in order from the pattern a. 0, 1, 1, the number of connections Ta ′ including the diagonal direction is 1, the number of connections Tb in the vertical and horizontal directions with peripheral pixels is 4, and the number of connections Tb ′ including the diagonal direction is a pattern. From 8, a, 8, 8, 10, 10,
8 and 8. In each of these six types of patterns a to f, the number of peripheral pixels p in the vertical and horizontal directions is 8, and the number of peripheral pixels p ′ including the diagonal direction is 12. Substituting these values into the above equations 3 and 4, z = (1 × 4) / 6 + (4/8 × 6) / 6 = 7/6 z ′ = (1 × 6) / 6 + (8/12 X4 + 10 / 12x2) / 6 = 31/18 is obtained.

Further, when the number of pixels is 3, in each of the four types of patterns a to d shown in FIG. 9, the number of connections Ta in the vertical and horizontal directions in the small matrix is 2 in each case.
The number of connections Ta ′ including the diagonal direction is 3, the number of connections Tb in the vertical and horizontal directions with peripheral pixels is 6, and the number of connections Tb ′ including the diagonal direction is 11. In each of these four types of patterns a to d, the number of peripheral pixels p in the vertical and horizontal directions is 8 and the number of peripheral pixels p ′ including the diagonal direction is 12. Substituting these values into Equations 3 and 4 above, Is obtained.

In addition, when the number of pixels is 4, in the pattern shown in FIG. 9, the number of connections Ta in the vertical and horizontal directions in the small matrix is 4, and the number of connections T including the diagonal direction is T.
a ′ is 6, the number of connections Tb in the vertical and horizontal directions with peripheral pixels is 8, and the number of connections Tb ′ including the diagonal direction is 12.
In this pattern, the number of peripheral pixels p in the vertical and horizontal directions is 8 and the number of peripheral pixels p'including the diagonal direction is 12. Substituting these values into Equation 1 and Equation 2 above, Is obtained.

The small matrix level connection probabilities z and z'determined as described above are stored in advance for each number of pixels in the small matrix, and are included in the large matrix M in the image to be image-processed. Number of pixels of each of the small matrices m00 to m33 Snm00 to Sn
The connection probabilities z and z'corresponding to m33 are read, and the sums of the read connection probabilities z and z'are set as the connection probabilities Z and Z'at the large matrix level.

For example, when the number of pixels in the set large matrix M is 24, in the pattern shown in FIG. 10A, Z = (1/4 × 8) + (7/6 × 8) ≈11.3. Z ′ = (5/12 × 8) + (31/18 × 8) ≈17.1. On the other hand, in the pattern shown in FIG. 10B, Z = (1/4 × 5) + (7/6 × 5) + (11/4 × 3) ≈15.3 Z ′ = (5/12 × 5) ) + (31/18 × 5) + (47/12 × 3) ≈22.4.

Therefore, even if the number of pixels of a particular density level belonging to the large matrix M centering on the target pixel A is the same, the connection probabilities Z and Z'are different depending on the pixel pattern. Therefore, the calculated connection probabilities Z and Z '
It is possible to determine the type of the image area to which the target pixel A belongs by comparing with the threshold value. For example, when the number of pixels in the large matrix M is 24, the connection probability Z
And a threshold T ′ for comparison with the connection probability Z ′ is 20, and if Z ≧ T and Z ′ ≧ T ′, the image type of the target pixel A located at the center of the large matrix M Is a character image and Z <T and Z ′ <
If T ', it can be determined that the image type of the pixel of interest A located at the center of the large matrix M is a halftone image.

In the image processing method according to this embodiment, 1 is set in the large matrix M of 8 × 8 pixels.
Since the six small matrices m are set without overlapping the pixels, when counting the number of pixels belonging to the sixteen small matrices m, the same pixel is not counted a plurality of times. Therefore, 16 small matrices m00 to m
The sum of the count values of the number of pixels in 33 can be directly used as the count value of the number of pixels in the large matrix M, and the calculation process of the connection probability can be simplified. In the image processing apparatus to which the image processing method of the present invention is applied, the first
The image processing can be executed for each pixel A according to the image type determined by any of the third embodiments.
For example, in the filtering process, the emphasis type filtering process is performed on the pixels determined to be the character image,
By not emphasizing pixels that are determined to be images other than character images such as halftone images, it is possible to clearly reproduce only character images and improve image quality.

Further, in the γ correction processing, the γ correction for clarifying the gradation is performed on the pixel determined to be the character image, so that the difference between the background portion and the character portion in the character image is clarified. Thus, the image quality can be improved.

Further, in the case of binarizing an image in an image processing apparatus such as a facsimile machine, for example, an input image having a density gradation of 8 bits including a halftone image and a character image is compared with a threshold value to obtain a binary image. When converting into a value,
By using the brighter value as the threshold value for pixels that are determined to be a character image, it is possible to prevent loss of character information, and by using the darker value as a threshold value for pixels that are determined to be a halftone image, photographs, etc. The image of is prevented from being crushed black. Further, the pixel determined to be a character image is binarized by comparison with a threshold value, and the pixel determined to be a halftone image is subjected to pseudo halftone binarization processing such as an error diffusion method. You can also do it.

In addition, in the image processing apparatus such as a facsimile apparatus, when the binarization processing is performed after changing the size of the multi-tone input image, for example, the resolution is 406.4 dp.
20 × 8-bit input image of i × 391.2 dpi
When transmitting in fine mode with a resolution of 3.2 dpi × 195.6 dpi, when the input image is reduced to 50% × 50% and then binarized and transmitted, the input image is 50% × 5.
While reducing to 0%, the first to
Similarly, the value of the connection probability calculated by the image processing method according to any one of the third embodiments is also reduced, and the above-mentioned 2 is determined based on the result of determining the image type of each pixel based on the reduced connection probability.
Quantization processing can be performed. This is the same when the filter processing or the γ correction processing is performed on the image after the scaling. In each of the above-described first to third embodiments, a method of discriminating the image type of each pixel in a multi-gradation image having 8-bit density data for each pixel has been described. Even when determining the image type of each pixel in a binarized image having value data, the present invention can be implemented in the same manner by calculating the connection probability for black pixels.

In an image processing apparatus such as a facsimile machine, the resolution may be different in the vertical and horizontal directions of the image. In this case, a display having the same vertical and horizontal resolution as the received image is received. However, if the image is displayed on the screen, the image may be displayed slenderly depending on the resolution of the received image, or the displayed size may change depending on the resolution of the received image even if the image size is the same. Therefore, in order to display the received image on the display in the same size as the transmitted image in the facsimile apparatus, processing such as resolution conversion or magnification conversion is required.

For example, the resolution in the main scanning direction is 203.2.
When a received image of a binary image having a resolution of 391.2 dpi in the sub-scanning direction is enlarged in the main scanning direction to obtain a square image, if the binary image is subjected to the magnification conversion processing, each dot becomes large. Even if a multi-valued process is performed after that, sufficient image quality cannot be obtained. Further, when the binary image is uniformly multi-valued by the smoothing filter process and the magnification conversion process is performed, the reproducibility of characters and the like deteriorates.

Therefore, based on the discrimination result of the image type of each pixel by the image processing method of the present invention, the character area of the binarized image is subjected to the emphasis filter processing, and then the other areas are subjected to the smoothing filter. By performing the magnification conversion process after performing the process, the image quality of the character image can be maintained in a good condition.

In the case where the received binary image is scale-converted into a multi-valued image with a specific resolution at a predetermined scaling factor, the connection probability calculated for each pixel is stored in the image in correspondence with the pixel configuration. When the information plane is scaled with the same scaling factor as the image and area information is stored for each pixel of the same size as the pixel of the multivalued image after scaling, only the area of the character image is displayed. It is possible to easily perform processing such as displaying only the area of the character image in another color.

FIG. 11 is a view showing the arrangement of an image processing apparatus to which the image processing method according to the embodiment of the present invention is applied. The image processing device 10 includes an input device 11 to which a binary image is input, a determination device 12 to determine the image type of each pixel included in the input image input from the input device 11, and an input image after the image type determination process. Apparatus for performing multi-value processing on the image, a density conversion apparatus for performing density conversion processing on the multi-valued image obtained by the multi-valued apparatus, and magnification conversion for the multi-valued image after the density conversion processing. Magnifying device 1 for processing
5, a binarization device 16 that binarizes the multivalued image after the magnification conversion process, and an output device 17 that outputs the binary image obtained by the binarization device 16.

The discrimination device 12 executes the image processing method according to any one of the first to third embodiments on the input image which is a binary image, for each black pixel, and for each pixel included in the input image. While discriminating the image type of, the connection probability for each pixel calculated in the image type discrimination processing is
A region information plane stored corresponding to the pixel configuration of the input image is created and supplied to the multi-value quantization device 13, the density conversion device 14, and the scaling device 15. Further, scaling device 15, the area information pre <br/> over emissions and magnification conversion processing using the scaling factor used in the magnification conversion processing of the input image, the area information plane after magnification conversion processing binarizing circuit Supply to 16. At this time, when the input image is enlarged, the connection probability of the pixel newly added in the image after the enlargement processing can be obtained by, for example, linear linear interpolation.

Further, considering that the information finally required for each pixel for performing image processing according to the image content is about three types of images, the amount of information for each pixel is relatively large. It can be said that there is little need to create a region information plane with a large number of connection probabilities. Therefore, the area information plane is configured by the image type that is the determination result for each pixel, and the area information plane configured by the image type for each pixel is scaled by the scaling factor of the input image. It is possible to reduce the amount of information to be stored in the memory, reduce the required memory capacity, and simplify the interpolation process. At this time, when enlarging the input image, the connection probability of the pixel newly added in the image after the enlarging processing can be obtained by a method similar to the nearest neighbor interpolation of the density data, for example.

The image processing apparatus 11 configured as described above
Then, the input image having the same size area information plane is subjected to filter processing and γ correction processing based on the area information to create a multivalued image, and the multivalued image is adjusted to the resolution of the output device 17. Then, the resolution is converted, and finally a binary image is output. The scaling device 15 receives the resolution information of the input image together with the input image, and determines the scaling factors Rh and Rv in the main scanning direction and the sub scanning direction when creating an image of the same size as the input image at the resolution of the output device 17. calculate.

As described above, the scaling device 15 performs scaling conversion processing on the multi-valued image and the area information plane by the scaling factors Rh and Rv in the main scanning direction and the sub scanning direction, respectively. For example, the input image is 406.4 dpi.
If the resolution of the output device 17 is a binary image having a resolution of × 391.2 dpi and the resolution of the output device 17 is 600 dpi × 600 dpi, the scaling factor Rh at the scaling factor conversion processing in the scaling device 15 is performed.
Is 147.64%, and the scaling factor Rv is 153.37%.

When the binarization device 16 binarizes the multivalued image after the magnification conversion processing based on the information of the area information plane after the magnification conversion processing, the binarization is performed for the pixels belonging to the character image. The threshold value is set low to prevent the loss of characters and ruled lines, and the pixel belonging to the halftone image is set to a high threshold value for binarization to prevent the image from being blackened. As a result, the image quality of the image output from the output device 17 can be maintained good.

In the binarization processing of the multivalued image after the magnification conversion processing in the binarization device 16, the binarization threshold is set low for the pixels belonging to the character image to prevent the loss of characters and ruled lines. In addition, the pseudo-halftone binarization process can be performed on the pixels belonging to the halftone image. Thereby, the reproducibility of the halftone image in the image output from the output device 17 can be improved.

[0081]

Effects of the Invention According to the invention described in claim 1, attention
A plurality of second pixels set in the first area centered on the pixel
A series corresponding to the same number of pixels in each area
The image type of the pixel of interest is determined by comparing the sum of the connection probabilities with a threshold.
By separating, the first area centered on the pixel of interest
Connection accuracy calculated based on multiple pixel patterns in
The image area of each pixel of interest can be determined from the
It is possible to improve the discrimination accuracy of the image area.

According to the second aspect of the present invention, when each pixel of the input image has multi-tone density information, the image of the pixel included in each of all or some of the plurality of density ranges is displayed.
By determining the image type of the target pixel based on the sum of the connection probabilities corresponding to the prime numbers, even when the input image is a multi-valued image, the target pixel is selected from the plurality of pixel patterns in the first region including the target pixel. The image type can be easily discriminated.

According to the invention described in claim 3, when each pixel of the input image has binary density information, for example,
Corresponds to the number of black pixels in the area including the pixel of interest .
By determining the image type of the pixel of interest based on the connection probability , the image type of the pixel of interest can be easily determined from a plurality of pixel patterns in the first region including the pixel of interest even when the input image is a binary image. Can be determined.

[0084]

According to the invention described in claim 4 , the first value calculated based on the result of counting the number of pixels in the second area
By determining the threshold value used to determine the image type of the pixel of interest from the number of pixels in the area,
It is not necessary to count the number of pixels in the area, and the result of counting the number of pixels in the second area can be effectively used, and the image type determination process can be further simplified and shortened.

According to the fifth aspect of the invention, the pixel of interest is calculated from the connection probabilities calculated based on the respective pixel patterns of the plurality of second areas set without overlapping the pixels in the first area. By determining the image type, the second number of regions can be efficiently set in the input image, and the region referred to in determining the image type can be enlarged to further improve the determination accuracy.

According to the invention described in claim 6 , when the magnification conversion processing is performed on the input image, the information plane formed by the connection probability calculated for each pixel of the input image is set to the input plane of the input image. By performing conversion with the same scale factor as the scale factor in the scale factor conversion process, the same scale factor as the scale factor in the scale factor conversion process of the input image at the time of the scale factor conversion process of the input image in resolution conversion etc. accompanying the adjustment of the output image size. It is possible to accurately determine the image type of each pixel after the magnification conversion processing by using the connection probability for each pixel that constitutes the information plane after being converted by.

According to the invention described in claim 7 , when the magnification conversion process is performed on the input image, the information plane formed by the image type previously determined based on the connection probability for each pixel of the input image. Is converted with the same scaling factor as the scaling factor in the scaling process of the input image,
During the magnification conversion processing of the input image in the resolution conversion and the like accompanying the adjustment of the output image size, the magnification conversion processing of the information plane configured by the value that identifies the image type having the extremely small amount of data compared to the connection probability of each pixel is performed. For,
It is possible to reduce the burden of data storage and calculation as compared with the case where the magnification conversion processing is performed on the information plane configured by the connection probability of each pixel.

[Brief description of drawings]

FIG. 1 is a flowchart showing a procedure of an image processing method according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating an image processing method according to the first embodiment of the present invention.

FIG. 3 is a diagram showing the relationship between the number of pixels in a small matrix and the connection probability in the image processing method according to the first embodiment.

FIG. 4 is a diagram illustrating a process of discriminating an image type of a pixel of interest by the image processing method according to the first embodiment.

FIG. 5 is a diagram illustrating an image processing method according to a second embodiment of the present invention.

FIG. 6 is a diagram showing the relationship between the number of pixels in a small matrix and the connection probability in the image processing method according to the second embodiment.

FIG. 7 is a diagram illustrating a process of discriminating an image type of a pixel of interest by the image processing method according to the second embodiment.

FIG. 8 is a diagram illustrating an image processing method according to a third embodiment of the present invention.

FIG. 9 is a diagram showing the relationship between the number of pixels in a small matrix and the connection probability in the image processing method according to the third embodiment.

FIG. 10 is a diagram illustrating a process of discriminating an image type of a pixel of interest by the image processing method according to the third embodiment.

FIG. 11 is a diagram showing a configuration of an image processing apparatus to which the image processing method of the present invention is applied.

[Explanation of symbols]

10-Image processing device 11-input device 12-discriminating device A-target pixel G-Image to be processed (input image) M-Large matrix (first area) m-small matrix (second area)

─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Mihoko Tanimura 22-22 Nagaikecho, Abeno-ku, Osaka City, Osaka Prefecture Sharp Corporation (56) References JP-A-2-292957 (JP, A) JP-A-4- 35167 (JP, A) JP 5-191642 (JP, A) JP 5-63970 (JP, A) JP 9-251537 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) H04N 1/40-1/409 H04N 1/46 H04N 1/60

Claims (7)

(57) [Claims] 1. A centered on the pixel of interest is an image type determination target
The first area composed of a predetermined number of pixels is set in the input image, and the first area is defined in the first area.
The number of pixels is less than the number of pixels
A plurality of second areas are set, the number of pixels of the same type in each of the second areas is counted, and the number of pixels corresponding to this count result is set.
The sum of the connection probabilities for each of the second regions
Predetermined for each number of pixels of the same type in one area
An image processing method characterized in that the image type of the pixel of interest is determined by comparing it with a threshold value . 2. An input image having multi-tone density information for each pixel is divided into a plurality of ranges of multi-tone density information,
Pixels included in each of all or part of multiple ranges
Of the connection probabilities for each of the second regions
The sum is calculated in advance for each number of pixels of the same type in the first area.
The image processing method according to claim 1, wherein the image type of the pixel of interest is discriminated by comparing with a predetermined threshold value . About wherein the input image having a density binary information for each pixel, communication for each of the second region about the pixel with one of the density information of the two values
The sum of the connection probabilities is the number of pixels of the same type in the first area
The image processing method according to claim 1, wherein the image type of the pixel of interest is discriminated by comparing with a predetermined threshold for each . 4. The number of pixels counted for the second area
Calculate the number of pixels in the first region based on the sum,
From this calculation result, the threshold to be compared with the sum of the connection probabilities
The image processing method according to claim 1, which is defined. 5. Each of the second regions is a first region
The image processing method according to claim 1 , wherein the pixels are set without overlapping in each other . 6. When the magnification conversion processing is performed on an input image, the
Information plane composed of connection probabilities for each pixel of interest
Magnification conversion processing with the same scaling ratio as the input image
Concatenation for each pixel that constitutes the information plane after magnification conversion processing
Determine the image type of the pixel of interest based on the probability,
The image processing method according to claim 1 , wherein processing of content according to a result is executed . 7. When the magnification conversion processing is performed on the input image,
The image type determined for each pixel of interest based on the connection probability
Therefore, the information plane constructed by the
Magnification conversion processing is performed with a single scaling factor, and the information
Processing of contents according to the image type of each pixel forming the lane
The image processing method according to claim 1 , further comprising: