JPS6031384B2 - Image processing method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an image processing method for restoring an image including halftones by binary output.

Generally, in image input/output devices such as digital copying machines and facsimile machines, when transmitting image signals read on the input side to the output side, the transmission efficiency is improved.
Also, when it is necessary to store the data contents, a big problem is how to reduce the storage capacity and restore an image close to the original image on the output side with binary output. ing.

Conventionally, images containing halftones such as photographs are scanned pixel by pixel.
As a method of processing the image signal obtained by sampling and restoring a high-gradation image with binarized output,
There is a so-called systematic dithering method.

This means that the entire image consisting of a matrix pixel array is n
It is covered with a submatrix consisting of ×n configuration, and a predetermined dither pattern is prepared for that submatrix.
The black or white level of each pixel is determined using the dither value as a threshold. However, in the conventional crude dither method, an image including halftones is sequentially scanned pixel by pixel using an image sensor, and the resulting image signal is quantized by n-values using an A/D converter. Even if the above-mentioned processing is performed according to each density level of the converted image signal and the image is restored by the binarized output,
The number of gradations of the restored image cannot exceed n.

For example, if the image signal is quantized on the input side and has 4 levels, it will be restored on the output side, but the image will be less than 4 gradations, and conversely, a restored image with 8 gradations will be required on the output side. When the image has 4 gradations, it is necessary to perform 8-value quantization on the input side. Therefore, when the image has 4 gradations, a 2-bit A/D converter can be used, but when the image has 8 gradations, it can be done with 3 bits, and even 1-bit quantization. In this case, each 4-bit A/D converter is required, and the more gradations there are, the more A/D converters are required. At the same time, there is a drawback that the more the quantized signal on the input side becomes multi-valued, the more bits are required to send the quantized signal to the output side, resulting in poor transmission efficiency and longer transmission time. The present invention has been made in consideration of the above points, and it reduces the number of bits of the quantized signal by quantizing an image including halftones at a low gradation, and also reduces the number of bits of the quantized signal. The present invention provides an image processing method capable of restoring high-gradation, high-quality images through binary output.

The image processing method according to the present invention takes advantage of the property that in images containing halftones, there is a very high density correlation between a pixel of interest and its surrounding pixels. The predicted density level of the pixel of interest is calculated from the quantized density level of the low gradation using a specific model formula that takes into account the modulation coefficient (magnification of gradation correction from low to high), and the restored image is calculated. Select one from a plurality of threshold patterns prepared in advance according to the number of gradations required for the process, and combine the threshold corresponding to the pixel of interest within the selected threshold pattern with the calculation result. By making a comparison, a binary determination of the black level or white level of the pixel of interest is made.

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

Figure 1 shows a specific pixel area with a 3x3 configuration in which the image is sampled pixel by pixel, and each pixel Pi-,,j
FIG. 2 shows the density level of each pixel when -, ~Pi-, and j- are quantized at low gradations.

In the image processing method according to the present invention, first, the pixel of interest Pi,
Density level Q of j:,j and its surrounding 8 pixels Pi-,,i
−・〜Pi−・, i, Pi+,, i〜PM core, each concentration level Qi−,,i−,〜Qi−,,i,QM,jYQ
i+, , j+, and the predicted density level li,j of the pixel of interest Pi,j is calculated according to the following equation.

1M=k(a・Qiーー, j・10b・Qi, j−・10C・Qi+1, j10・10d・Qi−1, j+e・Qke10f・Qi+,, i+g・Qi−・,i ten・tenh・Qi,
i ten・ten i・Qi+,,M)...
[1} Here, k is a modulation coefficient, which is given by the number of output gradations/the number of input quantizations, and a to i are weighting coefficients corresponding to each pixel, a + b + c + d + e + f
+g+h+i=1.

Next, prepare a plurality of threshold patterns (see Figure 4) in which the optimal threshold corresponding to each pixel is set in advance according to each value of modulation coefficient k, and select the required value of modulation coefficient k. One threshold pattern Tk is selected according to.

Then, the selected threshold pattern Tk
Threshold t, blood corresponding to the pixel of interest in is extracted, and its value t,,m is compared with the calculation result li,j, and the black or white level of the pixel of interest is determined according to the following determination result. let {i) 1w and t, when it comes to skin“
1" (bad level) output 'ii' When li,i<tl,m By quantizing the tone, it is possible to restore a high-quality image with high gradations from a degraded image.

As an example, the restored image is obtained by quantizing the density information of each pixel obtained by scanning an image including halftones, converting the quantized signal into density information of one Naga tone, and then binarizing it. The case of obtaining is explained below.

Note that FIG. 3 shows an example of the density level state of a four-level quantized image. Also, in this case, the modulation coefficient k is k=1
6/4=4, and FIG. 4 shows a threshold pattern in which an optimal threshold is set. Now, to simplify the explanation, the pixel of interest Pi,j and
Among the surrounding pixels, there are four pixels Pi, j-, , Pi- . ,i,Pi+・,i,P
The predicted density level 1;, j of the pixel of interest is determined using each density level i, i+.

Furthermore, if each weighting coefficient is set to a constant a, then a=0.2 from b+d+e+f+h=1, and according to the above equation {1}, the predicted concentration level li,j in this case can be obtained by the following equation. Become. IM=4×(〇.2XQ;
,H+〇. 2XQH,j+0.2×Qi,j+0.2×
Q River, j 10.2 x Qi, j+,)... ■For example, at the density level of each pixel with the four gradations shown in Figure 3, A in the figure
Focusing on the density level of each pixel in the 3×3 pixel area shown by Level Qi,j-,,Qi,i+
・, Qi-・,i, Qi+・,i are 1, 2, 2, respectively
, 2, so by substituting those values into the above equation (■), we get
Pixel of interest P when converted to image information based on Ichiro-kyo tone:
, j is calculated as 6 (
(Round to the nearest whole number.)

The predicted density level li of the pixel of interest calculated in this way,
By comparing j with the threshold corresponding to the pixel of interest in the threshold pattern of FIG. 4 as described above, the output level of the pixel of interest is determined, that is, 2
It can be converted into value.

In this case, a binary restored image obtained by processing the image in FIG. 3 is shown in FIG. For example, if the density level 6 of the pixel of interest Pi,j at the center of the pixel area A in FIG. 3 is binarized using the threshold pattern shown in FIG. ,j
Since the threshold value corresponding to the position is 11, 6<11, which means that it is binarized as a white level. In Fig. 5, 2 corresponding to the pixel of interest P;,j
A pixel that has been converted into a value is indicated by a. Note that each element in the threshold pattern with a 4x4 matrix configuration shown in Figure 4 corresponds to the size of a pixel, and the pixel area in Figure 3 is covered with the matrix around 1l to generate a binary value. It is becoming more and more processed. FIG. 6 shows an example of a configuration for concretely implementing the image processing method according to the present invention, in which input pixel signals sequentially obtained by scanning an image including an intermediate section are input to a quantizer (A). - D converter) 1, where the density information is quantized at a low gradation level, and the quantized signal is sent to the storage device 2, where the density level information in a specific pixel area is temporarily stored collectively.

Next, in accordance with the contents of the memory 2 and the modulation signal k' applied from the outside, the arithmetic unit 3 calculates the predicted density level li,j of the pixel of interest based on the {li formula described above. In addition,
At this time, the address indicator 4 reads the contents of each pixel necessary for the calculation from the memory 2 in response to an address signal applied from the outside. At the same time, the modulation signal k is applied to the threshold pattern selector 5, where a pattern selection according to kl is performed, and the selection signal is sent to the ROM 6 in which a plurality of threshold patterns are stored in advance. In response to a command from the address indicator 7 for a threshold pattern corresponding to the address signal and the modulation signal k', the threshold corresponding to the pixel of interest is read from the ROM 6 from the selected pattern. Next, the output li,j of the arithmetic unit 3 and the output t skin of the ROM 6 are compared by the comparator 8, and a binary output pixel signal with a black or white level determined according to the magnitude determination result is obtained. It is configured as follows. Therefore, in the image processing device configured in this way, since low gradation images are read on the input side, the bit capacity of the quantizer 1 and the memory 2 can be reduced, and the quantized signal When it is necessary to transmit the data to the output side, there is an advantage that the transmission efficiency is improved and processing can be performed at high speed.

As described above, in the image processing method according to the present invention, an image including halftones is read by quantizing low gradations, and simple arithmetic processing is performed using the quantized signal, and high gradations are processed by binary output. It has the excellent advantage of being able to restore high-quality images.

[Brief explanation of the drawing]

Figure 1 is a diagram showing a specific pixel area in which an image is decomposed into pixel units, Figure 2 is a diagram showing density level information obtained by quantizing each pixel at low gradations, and Figure 3 is a diagram showing quantization of an image in four gradations. FIG. 4 is a diagram showing an example of the threshold pattern, and FIG. FIG. 6, which is a diagram showing a restored image, is a block diagram showing an example of a configuration for specifically implementing the image processing method according to the present invention. 1... Quantizer, 2... Memory device, 3... Arithmetic unit, 4
...Pixel address indicator, 5...Threshold pattern selector, 6...RO
M, 7…. . . . Threshold pattern address indicator, 8 . . . Comparator. Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6

Claims (1)

[Claims] 1 In a method of restoring an image containing halftones using a binary image, modulation is performed from the density level of each pixel in a specific pixel area centered on the pixel of interest, based on image information read at low gradations for each pixel. Calculate the predicted density level of the pixel of interest using a specific model formula that takes into account the coefficients and weighting coefficients of each pixel, and also calculate the required modulation coefficient value from among multiple threshold patterns prepared in advance. The optimum pattern is selected, and the threshold corresponding to the pixel of interest in the selected threshold pattern is compared with the predicted density level obtained by the calculation, and the pixel of interest is binarized. An image processing method characterized by: