JPH1155503A - Image processing unit - Google Patents

Abstract

PROBLEM TO BE SOLVED: To apply proper processing even to an image where characters and a natural image are mixed without losing sharpness of the characters and gradation of the natural image in the case of applying resolution conversion to a digital image. SOLUTION: Let a ratio of resolution conversion be an integer ratio of M:N, an original image is divided into a minutely divided area of M×M pixels and when a kind of a color components of pixels included therein is >= a threshold level, it is discriminated to be a natural image and image processing by a linear interpolation circuit 104 is conducted and in the case that the kind is <= the threshold level, it is discriminated to be a character image, and the image processing by an adaptive interpolation circuit 105 is conducted depending on a correlation coefficient for each of M×M pixel area divided minutely as above and the image is reduced into N×N pixels.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus, and more particularly, to an image processing apparatus having a feature in thinning means for converting resolution of an image.

[0002]

2. Description of the Related Art Recently, images are often handled by computers, and images are often displayed on monitors and printers. At this time, some monitors and printers have various resolutions, and the resolution needs to be converted in order to properly output an image to be displayed by these devices. For example, a screen having a pixel number of 640 × 480 or 8
When displaying a screen having the number of pixels of 00 × 600 on a monitor, if the monitor is a monitor using a CRT, the resolution is not converted by scanning an electron beam in accordance with the number of dots of each. Each screen can be displayed on the monitor. However, when displaying an 800 × 600 screen on a liquid crystal monitor having 640 × 480 pixels, it is necessary to convert the resolution of the original image to match the number of pixels of the liquid crystal before displaying. As a resolution conversion method, a nearest pixel replacement method, a linear interpolation method, and the like are widely known.

In the nearest pixel replacement method, resolution conversion is performed by simply thinning out an image.
When converting from 0 to 640 × 480, processing can be performed by sequentially thinning out one pixel every five pixels in the vertical and horizontal directions. This method is simple and can be performed at a high speed. However, since the data is completely thinned out for each line, there is a problem that information is lost or discontinuous and the image quality is considerably deteriorated. In the linear interpolation method, pixels after resolution conversion are created by performing a linear operation. In this example, 5 × 5 pixels are used.
The resolution conversion from 800 × 600 to 640 × 480 can be performed by obtaining 4 × 4 pixels from the pixels of the above by interpolation. With this method, the lack of information can be suppressed considerably, and very good results can be obtained with natural images, but on the other hand, the amount of computation is slightly increased, and edges such as characters and graphic images are clear. When an image is processed, an intermediate portion of a character and a portion other than a character is generated at the edge portion by the interpolation operation of the edge portion,
There was a problem that characters were blurred.

Therefore, in the resolution conversion of an image, a suitable process can be performed by switching the resolution conversion method between the natural image portion and the character portion in the image. As this method, JP-A-4-199477 can be mentioned. This is a linear interpolation method that determines whether the pixel of interest is a natural image portion or a character portion from the pixel of interest in the original image and its neighboring pixels, and if it is determined that the pixel of interest is a natural image portion, can maintain gradation. The resolution conversion is performed by the nearest pixel replacement method that can maintain the resolution if it is determined that the character portion is detected.

[0005]

As described above, the conventional image processing apparatus has two problems. One is a method for determining a natural image portion and a character portion, and the other is a method for determining a character portion. That is, the closest pixel replacement method is used for resolution conversion.

In the conventional image processing apparatus, the difference between the maximum density and the minimum density in the vicinity of the pixel of interest is focused on the method of discriminating between the natural image portion and the character portion. If it is above the threshold value, it is regarded as a character part. This utilizes the property that a natural image often has a small density difference within a limited area due to its continuity. However, in reality, for example, in an image such as a human face, high and low densities are close to pixels in a natural image, such as a pupil and a white eye of an eye portion, and a white light in a black pupil. There is a problem that, when this is processed by the conventional technology, the eye portion is thinned out and an image without a pupil is generated in some cases.

Further, for resolution conversion of a character portion, a conventional image processing apparatus uses a closest pixel replacement method.
This method is to unconditionally thin out periodically determined lines. According to this method, for example, even in a character such as a Mincho type character in which a horizontal line has a thickness of only one line, that one line may be thinned out. As a result, there is a problem that the image quality is remarkably deteriorated and the determination becomes difficult.

SUMMARY OF THE INVENTION It is an object of the present invention to surely discriminate a natural image portion and a character portion in resolution conversion of an image, and particularly to perform a process for a character portion so as not to lose the characteristics of the character as much as possible.

[0009]

In order to solve the above-mentioned problems, an image processing apparatus according to the present invention associates the number of pixels of an original image and a reduced image to be converted with an integer ratio M: N, and The image is subdivided into M × M pixel regions, and a reduced conversion process is performed for each subdivided image region M × M to create an N × N reduced subdivided image. In the image processing apparatus that creates a reduced image by combining the segmented images N × N, the reduction process for each of the subdivided image regions is a process of thinning out MN lines in any of the row and column directions, Further, a correlation coefficient of a line to be thinned is calculated for each of the subdivided image areas, and a thinned line is determined according to the calculated correlation coefficient.

According to the present invention, when converting the resolution of an image, even if the image includes both a character and a natural image at the same time, the character portion and the natural image portion are surely separated from each other, and are suitable for each. By performing the resolution conversion, it is possible to obtain a good image after the resolution conversion.

[0011]

An image processing apparatus according to a first aspect of the present invention associates the number of pixels of an original image and a reduced image to be converted with an integer ratio M: N, and converts the original image to M × N. Each pixel area of M is subdivided, a reduced conversion process is performed for each subdivided image area M × M to create an N × N reduced subdivided image, and all of the reduced subdivided images N × In the image processing apparatus for creating a reduced image by combining N, the reduction processing for each of the subdivided image areas is processing for thinning out MN lines in any of the row and column directions, and the line to be thinned out. Is determined for each of the subdivided image areas.In addition to preventing the blurring of edges by not performing a linear operation between pixels, important information is lost by thinning out fixed positions, and image quality is reduced. Has the effect of preventing deterioration

Next, in the image processing apparatus according to a second aspect of the present invention, in the first aspect, in the subdivided M × M pixel area, a correlation coefficient of a row direction line is set to a value of an adjacent row. The number of pixels having the same color component of the adjacent pixels is counted, and the correlation coefficient of the column direction line is created by counting the number of pixels having the same color component of the adjacent pixels of the adjacent column. It is characterized by thinning out one of the lines with the highest correlation coefficient between each line in the direction.By thinning out only the lines with high correlation, the amount of information lost in the image is reduced. The deterioration of the image quality can be minimized, and the correlation can be easily calculated without performing an operation that takes a long processing time such as multiplication.

Next, the image processing apparatus according to claim 3 of the present invention calculates the maximum value of the correlation coefficient of the row direction line in the subdivided M × M pixel region. Thereafter, if the line of the correlation coefficient maximum value becomes the same as the thinned line of the subdivided pixel region adjacent to the left of the pixel region, the calculated correlation coefficient is weighted with a predetermined weight,
After calculating the maximum value of the correlation coefficient of the column direction line, if the line of the maximum correlation coefficient becomes the same as the thinned line of the subdivided pixel area adjacent to the pixel area, the calculated phase is calculated. The present invention is characterized in that relation numbers are weighted in a predetermined manner, and to prevent discontinuities in horizontal and vertical line segments that occur when lines to be thinned out are different for each of pixel regions arranged side by side horizontally and vertically. Can be done.

Next, the image processing apparatus according to claim 4 of the present invention associates the number of pixels of the original image and the reduced image with an integer ratio M: N, and converts the original image into an M × M image. In an image processing apparatus that subdivides each pixel region and performs a reduction conversion process for each subdivided image region M × M, the number of types of color components of pixels is counted for each subdivided pixel region, and the count is performed. If the value is smaller than a predetermined value, the first reduced image processing is performed, and if the value is larger, the second reduced image processing is performed to convert the subdivided M × M image area into an N × N image. It is characterized in that it is reduced and converted into an area. By separating a character portion and a natural image portion in an image, resolution conversion processing suitable for each can be performed.

Next, in the image processing apparatus according to a fifth aspect of the present invention, in the fourth aspect, the first reduced image processing is performed in a sub-divided M × M pixel area in a row direction line. The correlation coefficient is to count the number of pixels having the same color component of adjacent pixels in adjacent rows, and the correlation coefficient of the column direction line is to count the number of pixels having the same color component of adjacent pixels in adjacent columns. And thinning out one of the lines having the highest correlation coefficient between each line in the row direction and the column direction, and wherein the second reduced image processing is a linear interpolation processing. The resolution conversion for preserving the smoothness can be performed for the natural image portion, and the resolution conversion for preserving the sharpness can be performed for the character portion.

(Embodiment 1) An embodiment of the present invention described in claims 1 to 4 of the present invention will be described below with reference to FIGS.

This embodiment is an image processing apparatus for performing resolution conversion for displaying a display signal of 800 × 600 pixels on a liquid crystal monitor having 640 × 480 pixels. FIG.
In the figure, reference numeral 101 denotes a subdivided pixel area creation circuit. In this resolution conversion, since it is sufficient to convert every 5 × 5 pixel into 4 × 4 pixels, the original 800 × 600 pixel image is converted into 5 × 5 pixels.
Subdivision is performed for each × 5, and 160 × 120 subdivided pixel regions are created. Reference numeral 102 denotes a color component counter that checks the RGB components of each of the 5 × 5 pixels of interest, and
Pixels having the same color component are all of the same type, and it is checked how many types of color components exist. If all the pixels are different, the count value is 25, and if all the color components are the same, the count number is 1. An image discrimination circuit 103 discriminates whether the focused 5 × 5 pixel is a character portion or a natural image portion. In addition, almost the same result can be obtained even if the determination is made based on only one color component or luminance component without counting all the R, G, and B color components.

The image discriminating circuit 103 judges that the count value output from the color component counter 102 is a character portion if the count value is equal to or less than a predetermined value, for example, 5 and judges a natural image portion if the count value is larger than 5. . This is because in the case of a natural image, it is extremely unlikely that all of the color components will be the same even in adjacent pixels due to the influence of capture noise and the like. This is based on the fact that each image has the same color component for every background, character, figure, etc., and there are not so many types. When the image discriminating circuit 103 determines that the image is a natural image portion, the image is converted into a 4 × 4 pixel image by the linear interpolation circuit 104, and when the image discriminating circuit 103 determines that the image is a character portion, adaptive thinning is performed. 4 × 4 by circuit 105
It is converted to a pixel image. The processing method of the linear interpolation circuit 104 and the adaptive thinning circuit 105 will be described later. The 160 × 120 subdivided pixel areas converted into a 4 × 4 image by both processes are reorganized as an image by the image reorganization circuit 106, and sent to each pixel of the liquid crystal monitor as a 640 × 480 image. ,Is displayed.

Next, the linear interpolation circuit 104 performs a linear operation on 5 × 5 pixels to convert them into 4 × 4 pixels. As shown in the following equation, Aij (i = 1 to 5, j = 1) 5) through Cij (i = 1 to 5, j = 1 to 4) to Bij
(I = 1 to 4, j = 1 to 4). Here, when processing is performed vertically and horizontally, B1j = C1j B2j = (3 × C2j + C3j) / 4 B3j = (C3j + C4j) / 2 B4j = (C4j + 3 × C5j) / 4 where Ci1 = Ai1 Ci2 = (3 × Ai2 + Ai3) / 4 Ci3 = (Ai3 + Ai4) / 2 Ci4 = (Ai4 + 3 × Ai5) / 4. In the above equation, a 5 × 5 pixel is converted to a 4 × 4 pixel via a 5 × 4 pixel. Of course, a 4 × 4 pixel is converted to a 4 × 4 pixel via a 4 × 5 pixel. Similar processing can be performed.

Next, the adaptive thinning circuit 105 will be described with reference to FIG. FIG. 2 shows 5 × 5 pixels before processing and 4 × 4 pixels after processing them by an adaptive thinning circuit. In this figure, what is displayed as "ic" can be read as "ic" even after the processing, and the processing is performed without loss of information. This can be realized by thinning out the second column and the third row of 5 × 5 pixels. The rows and columns to be thinned are determined according to the following procedure. First, in the column direction,
The pixels in each column are compared with the right neighboring pixels, and the number of pixels having the same density value is counted. FIG. 2 shows the case of a density value of any of R, G, and B components. The count number in the first and second columns is 1, and the count number in the second and third columns is 5. , The count number of the third and fourth columns is 1, and the count number of the fourth and fifth columns is 3. Among them, the one with the largest count number, that is, the one with the highest correlation is the second
In this case, the number of counts is 5 and both are exactly the same, so that even if one of them is thinned out, there is no loss of information. Therefore, the second column is thinned out.

Similarly, in the row direction, each pixel is compared with the adjacent pixel below, and the third row and the fourth row having the highest correlation are compared.
The third row among the rows is determined as a thinned row. In this example, since the third row and the fourth row are completely the same, the result is the same regardless of which row is thinned. However, when determining which of the third row and the fourth row is thinned, By checking the correlation between the third row and the second row and the correlation between the fourth row and the fifth row and thinning out the row having the higher correlation, it is generally possible to further suppress the lack of information. In this case, the count value of the third and second rows is 3, and the count value of the fourth and fifth rows is 4.
Therefore, the correlation between the fourth row and the fifth row is higher, and the fourth row is determined to be a thinned row.

Hereinafter, a general processing procedure in the adaptive thinning circuit will be described with reference to FIG. FIG. 3 is a processing flowchart of the adaptive decimation circuit, and illustrates a pixel Ai before conversion.
A procedure for determining a row X and a column Y to be thinned out from j (i = 1 to 5, j = 1 to 5) is shown. 201 to 208 are procedures for determining a row X to be thinned, and 211 to 218 are procedures for determining a column Y to be thinned. In 201, i = 1 is set, and the correlation with the lower row is examined sequentially from the first row. At 202, Pi = 0 is set, and the count value is initialized. In 203, the correlation is checked for five pixels in order from the first pixel.
At step 204, the density values of the color components of the target pixel and the pixel adjacent to the target pixel are checked. 206
Is repeated 204 times five times. By the above processing,
The correlation coefficient between the lower adjacent row is included in Pi. 2 in 207
The processing of 02 or less is repeated four times, and the correlation coefficient in each row is obtained. At 208, a row having the largest value among the correlation coefficients is checked, and the row is determined as a row to be thinned.

The same applies to 211 and below. That is, j = 1 is set in 211, and the correlation with the right column is checked in order from the first column. At 212, Pj = 0, and the count value is initialized. In 213, the correlation is checked for five pixels in order from the first pixel. At 214, the density value of the pixel of interest and the pixel adjacent to it are examined, and if they are the same, the count value is incremented by 1 at 215, and if not, 215 is bypassed. 216 and 214 or less are repeated 5 times. By the above processing, the correlation coefficient between the right adjacent column is P
Enter i. In 217, the processing of 212 or less is repeated four times, and the correlation coefficient in each column is obtained. At 218, the column having the largest value among the correlation coefficients is checked, and the column is determined as the column to be thinned.

The thinning line can be determined as described above. However, since the line to be thinned is completely individually determined for each subdivided pixel region in this state, the line to be thinned is different in the adjacent pixel region and the line direction is different. If there is a line segment extending between pixel regions, a problem may occur that the line segment becomes discontinuous at the boundary of the pixel region. FIG. 4 shows an embodiment of an adaptive thinning circuit for preventing this.

FIG. 4 is a processing flowchart of another embodiment of the adaptive thinning circuit, and the difference from FIG.
1, 402, 411, and 412 are added, and the last used in steps 401 and 411
(X) and last (Y) are the row number of the subdivided pixel area adjacent to the left and the column number of the subdivided pixel area adjacent to the upper side. That is, after calculating the correlation coefficient Pi between each row, it is determined whether or not the row for which the current correlation coefficient is determined by 401 is a row thinned out in the subdivided pixel area adjacent to the left.
If so, at 402, the correlation coefficient is weighted by adding 0.5. Similarly, for the column, it is determined whether or not the column for which the current correlation coefficient has been obtained in 411 is a column thinned out in the subdivided pixel region adjacent to the upper portion. 0.5 is added and weighted. This weighting increases the possibility of thinning out the same row (or column) as the subdivided pixel region adjacent to the left (or above), thereby reducing the probability of discontinuity.

[0026]

As described above, according to the image processing apparatus of the present invention, it is possible to convert the resolution of an image in which a character and a natural image are mixed without losing the sharpness of the character and the gradation of the natural image. Processing can be realized by simple processing.

[Brief description of the drawings]

FIG. 1 is a block diagram of an image processing apparatus according to an embodiment of the present invention.

FIG. 2 is a diagram schematically showing a subdivided pixel area used in the image processing apparatus according to the embodiment of the present invention.

FIG. 3 is a flowchart of adaptive thinning used in the image processing apparatus according to the embodiment of the present invention.

FIG. 4 is another flowchart of adaptive thinning used in the image processing apparatus according to the embodiment of the present invention;

[Explanation of symbols]

 Reference Signs List 101 Subdivided pixel region creation circuit 102 Color component counter 103 Image discrimination circuit 104 Linear interpolation circuit 105 Adaptive thinning circuit 106 Image reorganization circuit

Claims (5)

[Claims] 1. An original image and the number of pixels of a reduced image to be converted are associated with an integer ratio M: N, and the original image is subdivided into M × M pixel regions. An image processing apparatus for performing a reduction conversion process for each M to generate an N × N reduced subdivided image and combining all of the reduced subdivided images N × N to generate a reduced image. The reduction process for each segmented image region is a process for thinning out MN lines in both the row and column directions, and determining a line to be thinned for each of the subdivided image regions. Processing equipment. 2. In the subdivided M × M pixel area, the correlation coefficient of the row direction line is obtained by counting the number of pixels having the same color component of adjacent pixels in an adjacent row, and calculating the phase of the column direction line. The relation number is created by counting the number of pixels having the same color component of adjacent pixels in an adjacent column, and thinning out one of the lines having the highest correlation coefficient between each line in the row direction and the column direction. The image processing apparatus according to claim 1, wherein: 3. In the subdivided M × M pixel region, after calculating the maximum value of the correlation coefficient of the row direction line, the line of the maximum correlation coefficient is adjacent to the left side of the pixel region. When the pixel area becomes the same as the thinned line of the pixel area, the calculated correlation coefficient is weighted in a predetermined manner, and the maximum value of the correlation coefficient of the column direction line is calculated. The image processing apparatus according to claim 2, wherein the calculated correlation coefficient is given a predetermined weight when it becomes the same as the thinned line of the subdivided pixel area adjacent to the area. 4. The original image and the reduced image are associated with an integer ratio M: N, and the original image is subdivided into M × M pixel regions. In the image processing apparatus that performs the reduction conversion process, the number of types of color components of the pixels is counted for each of the subdivided pixel regions, and if the count value is smaller than a predetermined value, the first reduction is performed. An image processing apparatus, wherein the image processing is performed by performing a second reduced image processing when the image processing is large, and the reduced M × M image area is reduced and converted into an N × N image area. 5. In the first reduced image processing, in a subdivided M × M pixel area, the correlation coefficient of a row direction line is determined by calculating the number of pixels having the same color component of adjacent pixels in an adjacent row. Counting, the correlation coefficient of the column direction line is created by counting the number of pixels having the same color component of the adjacent pixel of the adjacent column, and the line having the highest correlation coefficient between each line in the row direction and the column direction is generated. 5. The image processing apparatus according to claim 4, wherein one of the lines is thinned out, and the second reduced image processing is a linear interpolation processing. 6.