JPH0135541B2 - - Google Patents

Description

[Detailed description of the invention]

INDUSTRIAL APPLICATION FIELD The present invention relates to a predictive conversion encoding device for encoding image information in a facsimile device for transmitting images with gradations such as photographs, an image storage device for storing images, etc. It is. Conventional configuration and its problems As a coding method for quantized multi-tone images,
Traditionally, DPCM (Differential Pulse Code)
Modulation) method is known. This method uses surrounding pixels to predict the value of the pixel to be encoded, and encodes the difference between the original pixel value and the predicted value, but in order to perform encoding faithful to the original image, The number of codes to express the difference signal increases,
It is difficult to obtain high compression ratios. In particular, one pixel is 8 bits (256 gradations), such as in commercial photo transmission.
It is difficult to obtain a high compression ratio when transmitting a multi-gradation image expressed in .
On the other hand, in order to obtain a high compression rate, a method is used in which the difference is quantized with a small number of bits to reduce the number of codes, but this method has the problem that the quality of the reproduced image is greatly degraded. Purpose of the Invention In view of the drawbacks of the conventional art, the present invention makes it possible to obtain many gradations during commercial photographic transmission, minimize changes in image quality of reproduced images, and obtain a high compression rate. A predictive transform coding device is provided. Structure of the Invention The present invention includes a first prediction unit that predicts a pixel value of a predicted original pixel of a multi-level quantized multi-gradation image using a plurality of surrounding pixel values; a pixel density conversion unit that converts the predicted value of the predicted original pixel calculated by the first prediction unit into a density, the difference between the density of the predicted original pixel calculated by the pixel density conversion unit and the density of the predicted value; A density difference calculation unit that calculates the density difference, divides the density range that each pixel of the multi-level quantized multi-gradation image can take into a plurality of density regions, and predetermines a density difference constant for each density region. a density difference constant output unit that outputs the density difference constant corresponding to the density range to which the density of the predicted original pixel converted by the pixel density conversion unit belongs; a density difference calculated by the density difference calculation unit and the density difference; Compare the density difference constant output by the constant output unit, select the pixel value of the predicted original pixel or one of the predicted values of the predicted original pixel according to the comparison result, and convert the predicted original pixel. an image conversion means comprising a judgment selection section outputting as a pixel value; and a second prediction for predicting a converted pixel value output from the judgment selection section constituting the image conversion means, based on a plurality of surrounding pixel values. By providing a predictive encoding means comprising a section, and an encoding section that encodes the error of the predicted value obtained by the second prediction section,
This aims to achieve the above objectives. DESCRIPTION OF EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 shows a block configuration of an image system having a predictive transform coding device according to an embodiment of the present invention. In FIG. 1, reference numeral 1 denotes an image reading device, which scans, samples, and quantizes a multi-tone image such as a photographic original. 2 is an image conversion device that converts the quantized pixel signal;
3 is an encoding device that encodes the output image from the image conversion device 2; 4 is a transmitting device that transmits the code; 5 is a receiving device that receives the transmitted code; and 6 is a decoding device that restores the image from the code. , 7 is an image recording device for recording restored images. Although the system shown in FIG. 1 is an example of an image transmission system, the transmitting device 4 and the receiving device 5 may be used as a storage device and a reading device from the storage device. The configuration of the image conversion device 2, which is a main part of the predictive transformation coding device, will be described in detail below with reference to FIG. First, in the image conversion device 2, the image reading device 1
Image signal 210 scanned and quantized by
is stored in the two-pixel memory 21. The two-pixel memory 21 stores an original pixel to be converted (hereinafter referred to as a converted pixel) and an original pixel next to the converted pixel (hereinafter referred to as a next converted pixel). This image signal is a signal having a value (generally referred to as a luminance signal value) proportional to the reflectance of the image original when the image reading device 1 reads reflected light from the original. A converted pixel signal 211 and a next converted pixel signal 212 are output from the 2-pixel memory 21, and the selector 23
One is selected by The selector 23 selects the signal 2 by the selection signal 226 from the selection signal generator 31.
11 or signal 212. The selection signal generator 31 controls the selection signal 226 so that the converted pixel signal 211 is selected at the start of conversion processing for one pixel. Output signal 213 from selector 23 is input to concentration converter 24 . The concentration converter 24 converts the input value into concentration according to the following equation. d=-log ₁₀ (1-L/LMAX) (1) Here, LMAX is the maximum value of the read image signal. For example, when one pixel is quantized with 8 bits, the number is 255. L is the input value to the concentration converter 24. d is the concentration. Equation (1) is applied to an image reading device where the relationship between the document and the read value is a large value in the dark areas of the document and a small value in the bright areas, and a small value in the dark areas and a large value in the bright areas. In the image reading device 1, the following equation is obtained. d=-log ₁₀ (L/LMAX) ...(2) Also, in equation (1) or equation (2), the inside of the bracket is 0.
In this case, d is determined to be a sufficiently large finite value. Note that the concentration converter 24 includes a logarithm calculator, a subtracter,
It can be realized using a divider, but ROM
(read-only memory). In the case of an image in which one pixel is expressed by 8 bits, there are 256 types of luminance signal values, and therefore there are also 256 types of density values d. The density values are calculated in advance for 256 types of luminance signal values, and the values are stored in the ROM.
By storing the luminance signal value as the address input of the ROM, the density value d can be obtained as output data from this memory. In this case, the required storage capacity of the ROM is 256 words. The word length is determined by the effective digits of the concentration d. For example, it is determined that the effective digits of the concentration d are up to the second decimal place, and an integer value obtained by multiplying the concentration d by 100 is stored, and in equation (1), If the value of d when the value inside the bracket is 0 is set to 2.55 or 5.11, the word length of the ROM becomes 8 bits or 9 bits, and a density converter can be realized extremely easily with a small capacity ROM. Next, a prediction method in the image conversion device 2 will be explained. The pixel signal 229 converted by the converting device 2 is output to the encoding device 3, and is also input to the image memory 22 and stored therein. The image memory 22 stores two scan lines: the scan line being converted and the scan line converted one scan before. Four reference pixel signals are output from the image memory 22. Further, the relationship between the pixel X to be converted and these reference pixels is shown in FIG. In FIG. 3, 31 is the current scanning line, 32
is a scanning line that was scanned one scan earlier than the current scanning line. Pixels to which pixel X is converted, P', Q', R',
S' are reference pixels and correspond to output signals 215, 216, 217, and 218 from the image memory 22, respectively. Pixel T' is the next pixel to be converted after pixel X. The predictor 25 receives the four reference pixel signals and the two pixel signals 2 from the 2-pixel memory 21.
11,212, three types of predicted values X~'signal 2
19, t~' signal 220, and X' signal 211 are output. The values of these signals are given by the following equations. The value X~' of the signal 219 is ′) 255) (When p′+(r′−q′)>255) …(3) The value t~′ of signal 220 is t~′=X X+(s′−r′) X (x+ When (s'-r')<0) (When 0x+(s'-r')255) (When x+(s'-r')>255) ...(4) The value X' of signal 221 is X= t t-(s'-r') t (when t-(s'-r')<0) (when 0t-(s'-r')255) (t-(s'-r') ′) > 255) …(5) In equations (3), (4), and (5), p′, q′, r′,
s' are the values of the reference pixels P', Q', R', and S', respectively, x is the value of the converted pixel This is the value of pixel T. Signals 219, 220, 22
1 is selected by the selector 26 and output to the concentration converter 27. The selector 26 is controlled by a selection signal 227 from a selection signal generator 31. The selection signal generator 3 selects the signal 219 at the start of one pixel exchange process.
1 controls the selection signal 227. The concentration converter 27 is realized by a ROM like the concentration converter 26. The output density signal 214 of the density converter 24 and the output density signal 223 of the density converter 27 are input to a subtracter 28 for subtraction, and the absolute value of the result is output as a density difference signal 224. On the other hand, the density signal 214 of the original pixel is also input to the density difference constant output device 32. The density difference constant output device 32 has the divided density regions and a density difference constant determined in advance for each density region, and calculates the density difference constant corresponding to the density region to which the density signal 214 belongs and outputs the density difference constant signal 23.
Output as 0. An example of concentration range division is shown in Table 1. Note that Table 1 is an example in the case of four divisions.

[Table] For example, when the value of the density signal 214 of the original pixel is 1.00, a value of 0.10 is output as the density difference constant signal 230. Although the concentration difference constant output device 32 can be configured using a comparator, it can also be easily realized using a ROM. For example, with the effective digit of the concentration d set to the second decimal place, the value obtained by multiplying the concentration d by 100 is
By inputting the address of the ROM and storing a value obtained by multiplying the concentration difference constant by 100 at each address (that is, concentration), the concentration difference constant can be obtained immediately. If the original image is 8 bits/pixel, 256 words is sufficient for the ROM capacity. For example, in Table 1, 2 is stored for addresses 0 to 29, 5 is stored for addresses 30 to 94, 10 is stored for addresses 95 to 129, and 20 is stored for addresses 130 to 255. in this way
By using the ROM, the density difference constant output device 32 does not become complicated even if the number of divisions of the density region is increased. The density difference signal 224 is compared with the density difference constant signal 230 in the judger 29, and if the density difference is not constant, the judgment result is 0, and if it is larger than the density difference constant, the judgment result is 1 as the judger output signal 225. Output. The image quality of the image after conversion is affected by the number of divisions of the density range, the boundary values of the divisions, and the density difference constant. As a result of evaluating the image in this example, sufficient image quality can be obtained with each constant shown in Table 1. A decider output signal 225 from the decider 29 is input to the selection signal generator 31. A flowchart of the operation of the selection signal generator 31 is shown in FIG. 4, and the operation will be explained below. A terminal 450 indicates the start of the conversion process for one pixel, and a terminal 451 indicates the end of the conversion process for one pixel. After starting the conversion process for one pixel, in process 401, the selector 23 converts the selection signal 226 into the converted pixel signal 211.
The selection signal 227 is set to a value that causes the selector 26 to select the value X~' signal 219. As a result, the subtracter 28 calculates the difference in density between the value of the converted pixel X and the values X~', and the determination result by the determiner 29 is determined in determination 402. Judgment 402
If the determination result signal 225 is 0, that is, the density difference is less than or equal to the density difference constant, in step 403 the selection signal 228 is set to a value at which the selector 30 selects the signal 219, and the conversion process for one pixel is completed. in this case,
The selector 30 outputs the value X~' as the converted value of the converted pixel X to the encoding device. On the other hand, the judgment result signal 2
25 is 1, that is, when the density difference is larger than the density difference constant, the selection signal 226 is sent to the selector 2 in process 404.
3 is set to a value that selects the next converted pixel signal 212, and the selection signal 227 is set to a value that causes the selector 26 to select the value t~' signal 220. As a result, the subtracter 28 calculates the difference between the density of the next converted pixel T and the value t~', and the judgment result is determined by the judgment unit 29 in judgment 405.
will be judged. Judgment result signal 225 in judgment 405
is 0, that is, when the density difference is less than or equal to the density difference constant, in step 406, the selection signal 228 is set to a value at which the selector 30 selects the converted pixel signal 211, and the conversion process for one pixel is completed. In this case, the selector 30 sets the value x of the converted pixel X to the converted value of the converted pixel X, and outputs the encoder. Judgment result signal 225
is 1, that is, when the density difference is larger than the density difference constant, in step 407, the selection signal 226 is set to a value at which the selector 23 selects the converted pixel signal 211, and the selection signal 227 is set to the value X'( signal 221)
Set to the value of your choice. As a result, the subtractor 28
The difference between the density of the converted pixel X and the value
08 will be determined. If the determination result signal is 0 in judgment 408, that is, the density difference is less than or equal to the density difference constant, in processing 409, the selection signal 228 is set to a value that causes the selector 30 to select the signal 221, and the conversion process for one pixel is completed. do. In this case, the selector 30 outputs the value x' as the converted value of the converted pixel X to the encoding device. If the determination result signal 225 is 1, that is, the density difference is larger than the density difference constant, in step 410 the selection signal 228 is set to a value at which the selector 30 selects the converted pixel signal 211, and the conversion process for one pixel is completed. In this case, the selector 30 outputs the value x of the converted pixel X to the encoding device 3 as the converted value of the converted pixel X. Output image signal 229 output to encoding device 3
is also input to the above-mentioned image memory 22, and the values output after conversion are sequentially stored. The image signal output from the image conversion device 2 is encoded by the encoding device 3. Next, the configuration and operation of the encoding device 3 will be explained with reference to FIG. As shown in FIG. 5, the output image signal 229 from the image conversion device 2 is stored in the image memory 51. The image memory 51 also stores scanning lines necessary for prediction when a reference pixel is included in a scanning line scanned in the past when predictive encoding is performed. From the image memory 51, reference pixel signals 510 and 51 are used to predict the value of the pixel to be encoded.
1,512 is output to the predictor 52. An example of the prediction method of the predictor 52 will be described below using FIG. 6. FIG. 6 is a diagram showing the relationship between reference pixels for prediction and pixels to be predictively encoded. 61 is the current scanning line, 62 is 1 below the current scanning line
Scan This is a scan line scanned in the past. Pixel X is a pixel to be predictively encoded, and P, Q, and R are pixels scanned in the past and are called reference pixels. Pixel T is a pixel to be predictively encoded next to pixel X, and pixel S is a reference pixel used for prediction of pixel T. The predictor 52 uses the reference pixels P, Q, and R to set the value x shown in the following equation to the pixel.
It is output as a predicted signal 513 of X'. X= p p(r-q) p (When p+(r-q)<0) (When o<p+(r-q)255) (When p+(r-q)<255) ...(6 ) (However, p, q, r are the reference pixels P,
These are the values of Q and R. ) The prediction signal 513 is the predicted pixel signal 514
is subtracted by the subtracter 53, and the result is the difference signal 5
It is output as 15. That is, the value ΔX expressed by the following equation is output as the difference signal 515. ΔX=X−X′ (7) (However, x is the predicted value of pixel X.) Here, the relationship between the image conversion device and the encoding device will be explained. Equations (3) and (4) are each isomorphic to Equation (6), and in process 403 in FIG. When output, the predicted value x and the value x of pixel X match in equation (6). Furthermore, when the value of the original pixel is output from the image conversion device 2 as the value of the converted pixel X in the process 406 of FIG. The pixel value is
The value is expressed by equation (3), and coincides with the predicted value in equation (6) when the pixel T next to pixel X in FIG. 6 is predictively encoded. When the value x according to equation (5) is output from the image conversion device as the value of the converted pixel X in the process 409 of FIG. 4, when the pixel T is converted in FIG. The value is shown by formula (3),
(6) in the predictive encoding of the pixel T next to the pixel X in Figure 6.
Matches the predicted value of Eq. In this way, in this predictor, the input image signal is the converted image signal from the image conversion device 2, and the image conversion process is performed using the same prediction formula as in predictive encoding.
In equations (6) and (7), pixel X where x=p, r=q
The appearance probability of is higher than that of the original picture, so the difference value △
The probability of appearance when X is 0 increases. The difference signal 515 is encoded by the encoder 54 and a code 516 is output. The difference signal output from the subtracter 53 has 511 types of values, and is encoded by assigning, for example, a Huffman code to each value. At this time, as mentioned above, the difference value △X
Since the probability of occurrence of =0 is high, by shortening the code assigned to the difference value 0, the total number of codes of the encoded image can be reduced, that is, the compression rate can be increased. The output code 516 is the transmitter 4
and is transmitted to the decoding device 6 via the receiving device 5. The decoding device 6 converts the code into a difference value, performs prediction using the same prediction method as the predictive coding device, restores the image by adding the difference value to the predicted value, and stores the image in the image recording device 7. The data is then recorded and played back. The table below shows the code length per pixel when the image conversion according to this embodiment is performed and predictive encoding is performed, and the code length per pixel when the same predictive encoding is performed without image conversion. The figure shows the improvement rate for the original image and the S/N ratio of the image after image conversion with respect to the original image. The image in the data in the table below is a 256-gradation image in which one pixel is expressed with 8 bits. Further, the S/N ratio is a value calculated from the mean square error between the original image and the converted image.

[Table] As described above, according to this embodiment, there is no image quality deterioration in the encoding device 3 and the decoding device 6, and the image conversion device 2
Because the image quality is controlled by the number of density region divisions, the boundary values of the divisions, and the density difference constant determined for each density region, there is little deterioration in image quality when multi-gradation images such as photographs are encoded, transmitted, or stored. Moreover, compared to a system that does not perform image conversion, the number of codes can be transmitted or stored, and the transmission time and storage capacity can be reduced. That is, in this embodiment, a multi-tone image is converted,
In order to predictively encode the converted image, the original image is divided into multiple density regions, a density difference constant is assigned to each density region, and the same prediction as that used for predictive coding is made after the conversion. The predicted value and the value of the original pixel of the image to be converted are each converted to density, and the density difference is calculated as the density to which the density value of the original pixel belongs. An image in which the predicted value is used as the converted pixel value if it is within the range of the density difference constant assigned to the area, and the original pixel value is used as the converted value if it is outside the range of the assigned density difference constant. A converter, an encoder that predictively encodes the converted image output from the image converter, and a coder that predicts using the same prediction method as the encoder and uses this predicted value and the transmitted error signal. A decoding device for reproducing images is provided. Therefore, when converting a multi-tone image using the difference value of the read value between pixels, the value of the read pixel is a value proportional to the brightness of the original, and the value of the read pixel between pixels in the dark part of the image is a value proportional to the brightness of the document. (luminance signal value) and
Even if the difference value of the read value (luminance signal value) between pixels in the bright area is the same value, the density difference between pixels in the dark area is significantly different from the density difference between pixels in the bright area, and it is visually are not the same, and in image conversion based on such read values (luminance signal values), the image quality deterioration in dark areas is greater than the image quality deterioration in bright areas, resulting in an image with visually greater image quality deterioration. The image conversion method according to this embodiment uses the density difference after converting the value (luminance signal value) obtained by reading the image into density, and further,
The original image is divided into multiple density regions by taking advantage of the difference in visual perception between density differences in dark areas and density differences in bright areas, and deterioration of image quality in both dark and bright areas is achieved by dividing the original image into multiple density areas and assigning a density difference constant to each density area. Since the image conversion uses the same prediction method as in predictive encoding, the compression rate is increased to improve the prediction matching rate of the converted image compared to when predictively encoding the original image. It can be improved. Effects of the Invention As described above, the present invention makes it possible to obtain many gradations during commercial photo transmission, etc., with little change in the image quality of the reproduced image, and in addition, to obtain a high compression ratio. The effect is also great.

[Brief explanation of drawings]

FIG. 1 is a block diagram of an image transmission system having a predictive transform coding device according to an embodiment of the present invention, FIG. 2 is a block diagram of an image conversion device which is a main part of the device, and FIG. Fig. 4 is a flow chart of the operation of the selection signal generator, which is the main part of the image conversion device in Fig. 2, and Fig. 5 is a conceptual diagram showing the relationship between pixels to be converted and reference pixels. The block diagram in FIG. 6 is a diagram showing the relationship between pixels to be predictively encoded and reference pixels. 2... Image conversion device, 3... Encoding device, 2
3, 26, 30... Selector, 24, 27... Concentration converter, 25... Predictor, 28... Subtractor, 29
...Determiner, 31...Selection signal generator, 32...
Density difference constant output device, 51... Image memory, 52...
...Predictor, 53...Subtractor, 54...Encoder.

Claims (1)

[Scope of Claims] 1. A first prediction unit that predicts a pixel value of a predicted original pixel of a multilevel quantized multi-gradation image using a plurality of surrounding pixel values; a pixel density conversion unit that converts the predicted value of the predicted original pixel obtained by the first prediction unit into a density, the density of the predicted original pixel converted by the pixel density conversion unit;
a concentration difference calculation unit that calculates a difference between the predicted value and the concentration;
The density range that each pixel of the multilevel quantized multi-gradation image can assume is divided into a plurality of density regions, and a density difference constant is determined in advance for each density region, and the pixel density conversion unit a density difference constant output unit that outputs the density difference constant corresponding to the density range to which the converted density of the predicted original pixel belongs; a density difference calculated by the density difference calculation unit and the density difference output by the density difference constant output unit; a determination selection unit that compares the pixel value with a constant and selects one of the pixel value of the predicted original pixel or the predicted value of the predicted original pixel according to the comparison result, and outputs it as a converted pixel value of the predicted original pixel; a second prediction unit that predicts the converted pixel value output from the determination selection unit constituting the image conversion unit, based on a plurality of surrounding pixel values; and the second prediction unit. What is claimed is: 1. A predictive transform encoding device comprising: a predictive encoding means having an encoding section that encodes an error in a predicted value obtained by the prediction encoding means; 2. The first prediction unit constituting the image conversion means and the second prediction unit constituting the predictive encoding means are configured to calculate the pixel value of the predicted original pixel by determining the pixel value of the predicted original pixel. 2. The predictive encoding device according to claim 1, wherein a plurality of converted pixel values and a pixel value of an original pixel to be converted after prediction of the predicted original pixel are used. 3. Claim 1, characterized in that the first prediction unit constituting the image conversion means and the second prediction unit constituting the predictive encoding means perform prediction using the same prediction formula. predictive coding device.