JPS61194973A - Image processing method - Google Patents

Abstract

PURPOSE:To enable variable power of images without degrading image quality by changing the matrix size when multi-value-quantized digital image informa tion for each picture element is binary-coded by the partial matrix method based on the spatial filter processing. CONSTITUTION:During the process of binary-coding the digital image information of each picture element by the partial matrix method, the size of the sub- matrix is changed to vary the power. In other words, sub-matrixes of 2X2, 3X3 or so are cut out of the mother matrix of 8X8 for a magnification, whereas cut-out sub-matrixes have a size of 4X4 for the unmagnification (100% variable power), and digital image information in the region of a corresponding size is put together and binary-coded. Thus, images reduced or enlarged with a 25%-step magnification or 50%, 75%, 125%... can be recorded on the printer. Longitudinal and lateral magnification of images can be independently changed by independently selecting the longitudinal and lateral sizes of sub-matrixes cut out of the mother matrix.

Description

[Detailed Description of the Invention] Shigeru 1. The present invention relates to image processing that changes the magnification of an image, and particularly relates to image processing that performs magnification processing by binarizing image information regarding the density level of an original image read in pixel units. Regarding processing method.

Kazue Sarame Traditionally, magnification processing to enlarge or reduce an image is performed by operating an optical system when scanning the image surface of a document with a scanner and reading the image information pixel by pixel, or by adjusting the scanning speed of the image surface. However, such means for changing the magnification of an image using optical system processing cannot be applied to digital copying machines that read and record images in real time. It is possible, but it cannot be applied when the image is scaled after the scanned image information is binarized and stored in memory, and the data is processed for the purpose of image quality improvement, image editing, etc. Can not.

Conventionally, as a method for scaling an image through data processing based on image information read in pixel units, pixel data is extracted or added according to a certain algorithm that takes into consideration the density level state of surrounding pixels. There is a method called interpolation that allows you to do this. However, when applying this interpolation method, there is no problem as long as the scaling ratio is within a small range, but as the scaling ratio becomes large, the reliability of the data as scanned image information of the original image deteriorates and the reproduction becomes worse. The quality of the image will be degraded. At this time, if the target image is an image containing halftones such as a photograph, the reproduction of the halftones is particularly poor.

1 viper The present invention has been made in consideration of the above points.

When scaling is performed by data processing based on image information read in each pixel, even if the image contains halftones, the reproducibility of the halftones is not impaired, and the image quality is reduced over a wide range of scaling ratios. The present invention provides an image processing method that allows scaling of an image to be performed without changing the size of the image.

In order to achieve the objective, the present invention first changes the size of the matrix when digital image information that has been multivalued quantized on a pixel basis is binarized by a partial matrix method using spatial filter processing. This allows the scaling process to be executed by

Hereinafter, one embodiment of the present invention will be described in detail with reference to the accompanying drawings.

In the present invention, an image plane including halftones of a photograph or the like is sequentially scanned with light in a line by a scanner, and image information read pixel by pixel by a line image sensor such as a COD is converted into a multi-value quantized digital image by AD converting the image information. Image information 2
When performing valorization.

Among various methods that perform spatial filter processing using a threshold matrix that utilizes the integral effect of the visual system in order to improve the reproducibility of halftones, data of the pixel of interest to be processed and a threshold value are used. A general partial matrix method is used in which binarization is performed by associating elements of submatrices obtained by dividing a matrix (mother matrix) into several parts.

Figure 1 shows a 4 x 4 submatrix 5M-4 cut out from an 8 x 8 mother matrix expressing 82 gradations and quantized to 64 gradations (6 bits) using the partial matrix method. A principle diagram for binarizing the digital image information DB pixel by pixel is shown. Also, in Fig. 2, 811 to S2
The principle diagram when a 4×4 sub-matrix 5M-4 is cut out from an 8×8 mother matrix MM having a threshold value of up to 2 is shown.

However, when applying such a partial matrix method,
The basics of the gradation of digital image information is determined by the size of the mother matrix, and the larger the mother matrix, the greater the gradation. The resolving power of the binarized image is determined by the size of the submatrix, and the larger the submatrix, the higher the resolving power. In reality, after considering such gradation and resolution, a submatrix of an appropriate size is cut out from a predetermined mother matrix to the extent that noise in the high frequency range is not noticeable, and digital image information is binarized. I will make you do it.

For example, if the number of gradations in the digital image information is 64, and the reading density of the original image by the scanner and the recording density of the image by the printer are the same (for example, 400 dots/inch), then from the 8 × 8 mother matrix Cut out 4X4
In order to record a same-size image on the printer side based on data obtained by binarizing the digital image information obtained on the scanner side using the partial matrix method using the submatrix of The information must be grouped into 4x4 sub-matrices. In other words, the pixel density of digital image information may be 1/4 of the recording density on the printer side. Although there are various methods of combining digital image information into one submatrix, the most common method is to take an average value.

In this case, by changing the size of the submatrix and changing the pixel area in which digital image information is combined, it becomes possible to record a magnified image on the printer side.

The present invention allows scaling processing to be performed by changing the size of the submatrix when digital image information is binarized pixel by pixel using a partial matrix method.

In other words, under the conditions described above, if the magnification is 100%
%) Sometimes I try to cut out a 4x4 submatrix from an 8x8 mother matrix;
From the mother matrix of 8, 2 x 2.3 x 3, 5 x 5.6 x 6
, 7×7, 8×8°, . . . are cut out, and the digital image information in pixel areas of corresponding sizes is combined into one and then binarized.

50%, 75%, 125%, 150% respectively.

Images that have been reduced or enlarged in 25% increments such as 175%, 200%, . . . can now be recorded on the printer side.

In Fig. 3, 8x with thresholds from SLl to S22 is shown.
From the mother matrix MM of 8 to the submatrix 5M of 3×3
A diagram showing the principle of cutting out -3 is shown. Cutting out other submatrices of different sizes is performed in the same manner.

When changing the scaling factor by changing the size of the submatrix cut out from the mother matrix, the vertical and horizontal sizes of the submatrix extracted from the mother matrix are separately selected. This makes it possible to change the magnification of the image in the vertical and horizontal directions independently. For example, the vertical direction is 10
If the same magnification is 0% and the magnification is 125% in the horizontal direction, cut out a 4 vertical x 5 horizontal submatrix from the mother matrix,
The extracted submatrix may be used to perform binarization processing of the digital image information.

If you want to perform more detailed scaling, you can use the partial matrix method to change the size of the submatrix used to convert digital image information into binary values based on the processed data that has been scaled. Then, as a second step, detailed scaling processing is performed using an interpolation method.

For example, when scaling is performed in 1% increments by interpolation according to processed data that has been binarized by the partial matrix method, as shown in Figure 4, the binarized '0' or Based on pixel data with ``<100
Considering a pixel array of I'll add more. At this time, since the scaling process in 25% increments has already been performed in the first stage, the second stage is to fill in the gaps between the 25% scaling factors.
It is sufficient to consider changing the magnification in 1% increments up to 24%.

In other words, 1. For example, when performing a reduction with a scaling factor of 30%, first, as described above, the image is reduced to 50% by binarization processing using a 2 x 2 sub-matrix cut out from an 8 x 8 matrix 71~RIX. According to the processed data, pixels are thinned out at a rate of 1 pixel for every 5 pixels in the 100 x 100 pixel array shown in Figure 4 in order to further reduce the remaining 20% of the details by interpolation. Perform such processing. At that time, in the pixel array of FIG. 4, i and j are 5. n (n= 1 t 2.3*...
・.

20) may be thinned out. 1
Pixels are thinned out one by one up to 20%, but in that case it is necessary to thin out the pixels evenly within a 100×100 pixel array. For example, if the pixel t, j = 5 is thinned out at 1%, then each pixel it j = 5.55 is thinned out at 2%.

An example of the order in which pixels are thinned out in this case is expressed by the values of i+J: 5, 55, 30, 80, 20, 70°45.95,
10,60,35,85,15,65.40,90,2
5, 75, 50, 100.

In this example, 20 target pixel positions were determined from the LOOXloo pixel array and thinned out sequentially.
By determining the position at which pixels are to be thinned out for each percentage and thinning out pixels at equal intervals, more accurate reduction processing can be performed.

For example, when enlarging at a variable magnification rate of 104%, first, the data is processed to 100% equal magnification through binarization processing using a 4x4 sub-matrix cut out from an 8x8 matrix. In order to further reduce the remaining 4% of the details by interpolation, we added one pixel for every 25 pixels in the 10QX100 pixel array shown in Figure 4. let In this case as well, determine the positions to add pixels so that the added pixels are equally spaced. For example, the pixel at each position of i, j = 25.50, 75, 100 and the one before A pixel is added between each pixel. At this time, the order in which pixels are added is determined while maintaining uniformity, as in the case of reduction. for example.

The values of i and j are in the order 25, 50, 75, 100, or 1% is 25.2%, 75.3% is 33, 67.100.4% is 25°50.75, 10
Make it 0 and so on.

The content of a pixel added to a predetermined position is determined based on the data content of surrounding pixels.

There are various ways to make this determination, but the simplest method is to make the following determination based on the data content of two adjacent pixels.

That is, when adding pixel M25 between pixel A24 at position j=24 and pixel A25 at position j=25 as shown in FIG. Each pixel A24. If both A25 are 1111', pixel M25 is set to "'l", and pixels A24 . A
25 are both '0'', the pixel M25 is set to '0''. Also, pixel A24 is ``1'' and pixel A25 is ``0''.
, or pixel A24 is II O71 and pixel A25 is +
+114, the data to be prioritized is determined in advance. In that case, if priority is given to It I ++, the pixel M25 becomes '1''.

Also, when adding the pixel M25 between the pixel A24 at the position of 1=24 and the pixel A25 at the position of j=25, the two vertically adjacent pixels The data of the pixels to be added in between can be determined according to the data content of becomes a problem.

There are two possible methods for determining the pixels in this shaded area. The first method is to make judgments based on the data contents of the four surrounding pixels that are adjacent to each other in the vertical and horizontal directions only for the pixels in the shaded area. That is, (24,
24), (24,25), (25,24), (25,2
5) If there are 3 or more of the 4 pixels, it is II 1 pg.
or II O31, choose that one, and if there are two, choose the one you have decided to give priority to in advance. The second method is to first determine between j=24 and j=25, for example, without performing vertical and horizontal processing at the same time, and then i;
24 and 1=25. Thereby, the data content of the pixels in the shaded area can be determined without requiring any special processing.

Since the scaling process using the interpolation method in this second stage is originally designed to perform independent processing in the vertical and horizontal directions, it is easy to change the scaling ratio in the vertical and horizontal directions independently. .

Therefore, as mentioned above. If this method is used in conjunction with means for executing the first stage scaling by appropriately changing the vertical and horizontal sizes of the sub-matrix, it is possible to perform vertical and horizontal scaling independently in detail. In the image processing method according to the present invention, when digital image information that has been multivalued quantized on a pixel basis is binarized by a partial matrix method using spatial filter processing, it is possible to This system performs scaling processing by changing the size of the submatrix to be cut out, so even if the image contains halftones, it does not impair the reproducibility of the halftones, and the image quality can be improved over a wide range of scaling ratios. It has the excellent advantage of being able to optimally change the magnification of an image without reducing the image quality.

[Brief explanation of drawings]

Figure 1 is a diagram showing the principle of binarization processing using the partial 71-Risk method, Figure 2 is a diagram showing the principle of cutting out a 4x4 submatrix from a mother matrix, and Figure 3 is a diagram showing the principle of cutting out a 4x4 submatrix from a mother matrix. FIG. 4 is a diagram showing the principle of cutting out a submatrix of tooxto. 5 is a diagram showing a state in which a new pixel is added between pixels in the horizontal direction. FIG. 6 is a diagram showing a state in which a new pixel is added between pixels when 1=j=25. It is. BS...Digital picture information MM...Mother matrix
SM...Submatrix

Claims (1)

[Claims] 1. Multi-value quantized digital image information for each pixel is n^
When binarizing by the partial matrix method using submatrices cut out from an n x n threshold matrix to express two gradations, the size of the submatrix cut out from the threshold matrix is changed. An image processing method that takes measures to change the magnification of an image to be binarized. 2. By separately selecting the vertical and horizontal sizes of the sub-matrices cut out from the threshold matrix, the magnification of the image in the vertical and horizontal directions can be changed independently. An image processing method according to the above item 1.