JPH0476548B2 - - Google Patents

Description

[Detailed description of the invention]

(Industrial Application Field) The present invention relates to a multi-tone image encoding device. (Prior art) Conventionally, as a technology in this field, Kishimoto et al., "Block coding method for still images" (Transactions of Institute of Electronics and Communication Engineers, 79/1, Vol. J62-B, No. 1, P17 ~
24). This will be explained below. According to the technique, an image is first divided into square blocks of 4 pixels each in length and width (that is, 16 pixels in total). Next, in each block, the average luminance within the block is set as a threshold, and pixels larger than this threshold are approximated by the average value a0 of pixels larger than the threshold,
Pixels smaller than this threshold are approximated by the average value a1 of pixels smaller than the threshold. Furthermore, if the root mean square error between these approximate values and the true value is small, each pixel is set to the two average values a0,
Either a1 is approximated, or each pixel in the block is approximated only by (a0+a1)/2. If the root mean square error is large, it is subdivided into square blocks of 2 pixels each in the vertical and horizontal directions (that is, 4 pixels in total), and each pixel in the block is approximated by an 8-bit or 4-bit average value a0, a1. It was something to do. (Problems to be Solved by the Invention) However, in the above encoding method, first of all, the processing is complicated, and if executed by software, the processing time will be long, and if executed by hardware, the amount of hardware will be large. There was a fear that it would become large-scale. Second, the code is different from the Modified Huffman (MH) code or Modified Read (MR) code that is generally used when encoding binary quantized documents, etc., so it is difficult to communicate or store data. In doing so, it is necessary to encode the binary image and the multivalued image separately, and the encoded data also has the problem of having to control the row synchronization signal etc. separately. It was hot. The present invention eliminates the above problems, enables multi-gradation images to be encoded with a small number of bits without increasing noise, and with simple processing, and also uses MH codes or MR codes used for encoding documents, etc. The purpose is to provide an encoding device that can perform (Means for Solving the Problem) In order to solve the above-mentioned problem, the present invention provides an image encoding device that encodes an image in which the gray level of one pixel is quantized using a binary signal of 2 bits or more. means for calculating a predicted gray value of the pixel of interest using neighboring pixels; means for calculating a degree of change in gray in the vicinity of the pixel of interest using the neighboring pixels; and a plurality of curves having the same number of steps for quantizing the approximation error into a plurality of preset values (approximation differences). A curve in which the absolute value of the approximation difference with respect to the absolute value of the approximation error is relatively small, and
means for selecting a curve in which the absolute value of the approximation difference is relatively large with respect to the absolute value of the approximation error when the degree of change in density is large; and means for selecting the approximate difference of the target pixel based on the selected quantization curve and the approximation error. means for determining the pixel of interest from its approximate difference to one of a number of states corresponding to each step of the quantization curve; and means for converting the state of each pixel of interest into two or more binary images. The apparatus includes means for converting, and means for encoding each binary image using a binary image encoding method. (Function) According to the present invention, each pixel of a multi-tone image expressed by a binary signal of 2 bits or more per pixel is determined by the difference between its predicted value and the true gray value and the change in gray level in the vicinity thereof. Based on the degree of change, if the degree of change in shading is small, the quantization level is a small curve, and
If the degree of change in shading is large, the information is compressed using a curve with a large quantization level and converted to a binary image,
Furthermore, the binary image is encoded. (Embodiment) FIG. 1 is a block diagram showing an embodiment of the apparatus of the present invention. Here, an apparatus for encoding and decoding both binary images and multi-gradation images is shown. In the figure, 1a is an input terminal for a binary image signal in which one pixel of a facsimile transmission document or the like is expressed by one bit, 1b is an input terminal for a multi-gradation image signal in which one pixel is expressed by a plurality of bits, for example, 8 bits, and 2 is the above 1
An image signal (hereinafter referred to as a special image signal) obtained by approximately combining a multi-gradation image signal of 8 bits per pixel with a large number of binary image signals of 1 bit per pixel as described later.
3 is a multiplexer, 4 is a preprocessing circuit that converts
5 is a coding circuit, 5 is a modulation/demodulation circuit in facsimile communication, communication path, etc. or communication or storage means such as an optical disk storage device, 6 is a decoding circuit, 7 is a demultiplexer, and 8 is a circuit for converting the special image signal into 8 bits per pixel. a post-processing circuit for converting into a multi-gradation image signal, 9
a and 9b are output terminals. In the above device, a binary image signal such as a facsimile transmission document inputted from an input terminal 1a is sent to an encoding circuit 4 via a multiplexer 3, where it is encoded using well-known encoding, for example, CCITT recommendation T4.
Encoding is performed in accordance with a one-dimensional encoding method such as MH encoding or a two-dimensional encoding method such as MR encoding shown in FIG. The encoded signal is outputted to the decoding circuit 6 via the communication or storage means 5, decoded, returned to the binary image signal, and further sent from the demultiplexer 7 to the output terminal 9a. Further, the multi-gradation image signal inputted from the input terminal 1b is converted into the special image signal in the preprocessing circuit 2, and sent to the encoding circuit 4 via the multiplexer 3, where it is encoded with the same one-dimensional code as described above. Encoding is performed in accordance with an encoding method or a two-dimensional encoding method. The encoded signal is output to the decoding circuit 6 via the communication or storage means 5, decoded, and returned to the special image signal. The special image signal is further sent from the demultiplexer 7 to the post-processing circuit 8, where it is converted into the multi-gradation image signal and sent to the output terminal 9b. FIG. 2 shows details of the preprocessing circuit 2. In the figure, 21 is an adder, 22 is a prediction circuit, 23 is a quantization circuit, and 24 is a binarization circuit. A predicted value x calculated by the prediction circuit 22 for an 8-bit image signal x indicating the gray level value (luminance information) for one pixel (pixel of interest), for example, X among input multi-gradation image signals. The difference between
sent to. The prediction circuit 22 calculates a predicted value of the luminance information x using the reproduced luminance information a〓, b〓, c〓 of three pixels A, B, and C having the positional relationship shown in FIG.
Calculate x^ as follows. x^=a〓/2+(b〓+c〓/4...(2) Here, the reproduced luminance information a〓, b〓, c〓 means that after decoding, based on the conversion by this preprocessing circuit 2, Indicates the luminance information of pixels A, B, and C that will be reproduced by the post-processing circuit 8, and in the prediction circuit 22, the reproduced luminance information x〓 for the pixel of interest ^+d...(3) is calculated and stored for future prediction.The prediction circuit 22 also simultaneously determines whether the density (luminance) is changing rapidly or if the change is small in the vicinity of the pixel of interest X. In order to make a determination, find the degree of change in shading (brightness) V as follows: V = MAX (a〓, b〓, c〓) - MIN (a〓, b〓, c〓)...
…(4) Here, MAX(a〓, b〓, c〓), MIN(a〓, b〓, c〓
)
are terms whose values are the maximum and minimum values of the reproduced brightness information a〓, b〓, c〓, respectively, and the degree of change V is the reproduced brightness information of the three neighboring pixels A, B, and C.
The degree of change is measured by the difference between the maximum and minimum values of a〓, b〓, and c〓. The quantization circuit 23 inputs the degree of change V and the approximation error D, and outputs an approximation difference d according to the graph shown in FIG. The binarization circuit 24 converts the approximate difference d into states S0 to S4 based on Table 1.

[Table] Here, the approximate difference d is 50, 30, 25, 20, 15,
It can take 13 values: 7, 0, -7, -15, -20, -25, -30, -50, but the same value is calculated for the degree of change V during decoding, so states S0 to S4 If the degree of change V is known, the approximate difference d can be uniquely determined from Table 1 and the graph of FIG. The binarization circuit 24 converts the approximate difference d into the state in this way.
After converting into S0 to S4, a pseudo binary image is created. Figures 5a and 5b show the multi-tone original image in the above state.
This shows how to convert into four binary images using S0 to S4. In the figure, 210 is an original image, 220 to 250 are images each having the same number of pixels as the original image and each pixel having 1 bit, and the images 220 to 250 are all set to "0" as an initial state. If the coordinates of the pixel 211 in the original image 210 are (u1, v1) and the state of this pixel 211 is S1 as a result of the above processing, then the image 220
Above, set the point 221 at coordinates (u1, v1) to "1 (black pixel)"
shall be. Similarly, the coordinates (u2, v2) on the original image 210
If the state of the point 212 is S4, the point 252 of (u2, v2) on the image 250 is set to "1". In this way, pixels in states S1 to S4 are expressed by setting one pixel on images 220 to 250 as "1". Regarding the S0 pixel,
"1" is not given to any of the images 220 to 250, and identification is thereby made. Next, the images 220 to 250 are horizontally connected as shown in FIG.
A doubled binary image, that is, a special image signal S, is output. FIG. 6 shows details of the post-processing circuit 8. In the figure, 61 is an approximate prediction error recovery circuit, 62 is a prediction circuit, and 63 is an adder. The special image signal S is generated by the approximate prediction error reproduction circuit 61
The image is decomposed into 220 to 250 images, and each pixel is
It is determined which of the states S0 to S4 the state is in (however, if all of the images 220 to 250 are "0", it is determined to be the state S0), and the state S0 is determined.
~S4 and the degree of change V calculated by the prediction circuit 62 from the equation (4) above, the approximate difference d is determined based on the graph of FIG. 4 and Table 1. Furthermore, the adder 63 adds the approximate difference d and the predicted value.
x^ is added, the above formula (3) is calculated, and the reproduced luminance information x is calculated and output. When the encoding device described above was applied to an actual image, the following results were obtained. The original image is color matching chart “A”, one of the test charts of the Television Society.
Girl with Carnation” was filmed with a TV camera,
Using samples sampled with 512 horizontal pixels and 480 vertical lines, MH encoding was used as the encoding method, and an 11-bit synchronization signal was inserted for each row, resulting in an average of 0.96 bits per pixel. The SN ratio of the reproduced image at this time was 34.9 dB, and the mean square error was 0.015.
A quality sufficient for practical use was obtained. (Effects of the Invention) As explained above, according to the present invention, when quantizing the approximation error between the true gray value of the pixel of interest and the predicted gray value into a plurality of preset approximation differences, If the degree of change in the density is small, the absolute value of the approximation difference with respect to the absolute value of the approximation error is relatively small according to the degree of change,
When the approximation error is large, a quantization curve with a relatively large absolute value of the approximation difference relative to the absolute value of the approximation error is selected, so a small number of quantization steps generates granular noise in areas with small changes in shading. It is possible to quantize a multi-gradation pixel of interest without causing noise that appears to be changing in position in areas with large changes in shading, such as edges, etc. Based on the difference, the pixel of interest is determined to be one of a number of states corresponding to each step of the quantization curve, the state of each pixel of interest is converted to two or more binary images, and then two or more states are determined for each binary image. Since encoding is performed using a value image encoding method, the standardized encoding method for binary images can be used as is, and when configured with hardware, the amount of hardware is small, and , When configured using software, there are advantages such as the software is simple and the processing time is shortened.

[Brief explanation of the drawing]

Fig. 1 is a block diagram showing an embodiment of the device of the present invention, Fig. 2 is a block diagram showing details of the preprocessing circuit, and Fig. 3 is an explanation of the positional relationship between the pixel of interest and neighboring pixels used for its prediction. Figure 4 shows the approximation error D and the approximation difference d with the degree of change in shading V as a parameter.
Figures 5a and 5b are explanatory diagrams showing how a multi-tone original image is converted into four binary images, and Figure 6 is a block diagram showing details of the post-processing circuit. . 2... Preprocessing circuit, 4... Encoding circuit, 6...
Decoding circuit, 8... Post-processing circuit, 21... Adder, 22... Prediction circuit, 23... Quantization circuit, 2
4...Binarization circuit.

Claims (1)

[Scope of Claims] 1. In an image encoding device that encodes an image in which the gray scale of one pixel is quantized using a binary signal of 2 bits or more, the predicted gray value of the target pixel is calculated using neighboring pixels of the target pixel. means for determining the degree of change in density near the pixel of interest using the neighboring pixels; means for determining the difference (approximation error) between the true density value of the pixel of interest and the predicted density value; and the approximation error. From multiple curves having the same number of steps for quantizing into multiple preset values (approximate differences), if the degree of change in shading is small, the absolute value of the approximate difference is relative to the absolute value of the approximation error. means for selecting a curve with a relatively large absolute value of the approximation difference with respect to the absolute value of the approximation error when the degree of change in shading is large; means for determining an approximate difference of the pixel of interest based on the approximate difference; means for uniquely determining the pixel of interest into one of a number of states corresponding to each step of the quantization curve from the approximate difference; An image encoding device comprising: means for converting into two or more binary images; and means for encoding each binary image using a binary image encoding method. 2. The image encoding device according to claim 1, wherein the difference between the maximum gradation value and the minimum gradation value in neighboring pixels is used as the degree of change in gradation. 3 CCITT Recommendation T4 as a binary image encoding method
The image encoding apparatus according to claim 1, characterized in that the image encoding apparatus uses a one-dimensional or two-dimensional encoding method defined in . 4. As a means of converting a large number of states into a binary image, two binary images with the same number of pixels as the original image are concatenated one less than the number of states, and the difference between the gray value of the pixel of interest and the predicted gray value is the largest. 2. The image encoding device according to claim 1, wherein when the image is in a state other than the state where the number of images is low, pixels at positions corresponding to the original image of the image corresponding to this state are set as black pixels.