JPS63164574A - Decoding system - Google Patents

Abstract

PURPOSE:To decode encoded image data without deteriorating picture quality by decoding the encoded image data to compressed color information and the original information and smoothing the decoded information. CONSTITUTION:When the compressed encoded code is inputted, color information C1, C2 are decoded 24-bit color information by decoding part 104 and area information 102 is decoded into 16-bit area information. The decoded image data are temporarily stored in a line memory 105 and sent to a smoothing circuit 106 for smoothing the decoded information in the unit of 3X3 picture element matrix. A decoding part 3 has a look-up table LUT for decoding and the LUT decodes the color information C1, C2 each of which consists of 12 bits to 24-bit color information constituted of red, blue and green components each of which consists of 8 bits. The 12-bit area information is also decoded to 16-bit area information and two color information C1, C2 decoded based on the area information are selected and sent to the line memory 105.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a decoding method for decoding image data compressed and encoded into two color information and area information.

[Prior Art] Conventionally, one of the methods for compressing color image data is
A method has been proposed in which mxm image data blocks are described using two-color block codes. This is, for example, the color information (r+, gI, b+) of 16 pixels of 4×4 (i=
1 to 16) to two representative color information (71, g1.b1)
: γ2°g2', b2), the information on these two colors and the information on the area are treated separately, and each is encoded and transmitted using predictive encoding, MR encoding, etc.

However, when the image memory is reduced by such image data compression and image editing is performed on the image memory, since the aforementioned predictive encoding method and MR encoding method use variable length codes, The efficiency of editing operations such as searching for data at an arbitrary address within an image and editing data, etc., becomes extremely low, making it impractical.

[Problems to be Solved by the Invention] The present invention has been made in view of the above-mentioned conventional examples, and at the same time achieves a high compression rate and reduces the quality of encoded image data so that image editing can be performed efficiently. The purpose of the present invention is to provide a decoding method that allows decoding without decoding.

[Means for Solving the Problems] In order to achieve the above object, the image data compression method of the present invention has the following configuration. That is, a predetermined pixel matrix of color image data is divided into two pieces of color information at most,
A decoding method for image data in which a color region is determined based on the color information, and the color information and the color region are respectively compressed and encoded, and the method is based on the compressed color information and the color region. and a smoothing means for smoothing the restored information.

[Operation] In the above configuration, the compressed color information and color area are restored to the original information, and the smoothing means operates to smooth the restored information.

[Embodiments] Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

[Explanation regarding color information [Figs. 6 to 8] First, compression of image data will be described.

6(A) to 6(C) are explanatory diagrams of the encoding method in this embodiment, and FIG. 6(A) shows a state in which two colors C, , C2 straddle an m×m image area 60. . The same figure (B
) shows a state in which the image area 60 is divided into two pieces of color information C, , C2. Figure (c) shows the area of figure (B) in mx
It is a figure expressed by the binary data of m. When the original image changes gradually, representative color information CI, C2 and area information are extracted according to the degree of change, as described later.

The extracted two color information CI and C2 are, for example, r of three colors.
, g, and b values. That is, Cs = , (r+
, gr , b+ )C2=(r2 r 82 + b2
) Now, assuming that each r+g+b is represented by 8 bits, 24 pits are required to represent one color. However, in reality, even 24 bits are not necessary, and our experiments have shown that images can be reproduced with 12 to 16 bits without degrading image quality. This is because color C outside the area where colors can be identified does not affect the image quality due to human visual characteristics.

FIG. 7 is a diagram showing a lookup table (LUT) for compressing image data. Here, 24-bit color information (r+, gI, b+) represented by three colors r, g, and b is shown.
is composed of a ROM, etc., and is input as an address signal of the UT70, and 12 to 16 bit color information (r'+,
'+, b't) is read out as converted color information.

Through this reduction, color information is reduced to color coordinates within small solids created when the isochromatic space is equally divided by the discrimination resolution size, and the total number of colors is determined by the number of small solids.

Furthermore, the color range to be expressed differs depending on what the image used is expressed with. For example, if a CRT is used, the NTS
It is sufficient to divide the chromaticity coordinate range defined by C (this is a triangle determined from the characteristics of the phosphor) into the range of identification resolution, and if it is only used for printing with a printer, etc., it can be sealed with printing reproduction. Only the chromaticity range is required.

On the other hand, the area information shown in FIG. 6(C) is further compressed by vector quantization. For example, in the case of 4x4, this vector quantization method converts 16 bits of area information (216 states) into bits of area states (216 states) such that the amount of distortion is minimized.
A vector on the phase space of 216 is converted into a vector on the phase space of 2 using the method of converting to the state of k). When encoding such a vector quantizer, it is assumed that the output vector that is the shortest distance from the input vector (resulting in the least distortion) is given in advance as a code table.

This code table is created by analyzing image data in advance using a statistical method. In the case of 4×4 binary data,
Although there are 216 states, it was found that there is not so much existing image data and that 212 states is sufficient. Therefore, = 12, and the area information is also approximately 3/4 (12
/16).

As described above, the 4×4 shown by 60 in FIG. 6(A)
As shown in Figure 8, the color image information of
The code information is converted into a total of 36 bits of code information consisting of 2 bits of color information C'r and C2 and 12 bits of area information.

The compression ratio η at this time is becomes.

[Explanation of color and area determination when image data changes slowly (Figures 9 and 10)] Color data of mXm pixels (X Ill Y IJIZ IJ) (i, j
−1 to m). Here (X IJIY IJI
Z IJ) represents the chromaticity coordinate, γ.

g, b series, X, Y, Z series, U, V, W series, Uw.

v", w" series, L', u", V" series, Ll.

Color systems such as a" and b" systems can be used. Here, for simplicity, it will be expressed as a γ, g, b system.

The two color information and area information are determined by the following algorithm.

(1) Average within all pixels within mXm (11g.

b) and variance (σ1.2.σb). Immediate σ Chi, σ, = Σ (XIj-γ)2/N ■ at −E”(Y+j g)”/N J QB = X” (Xl”b) 2/N■ However, N=m'.

(2) Next, σ1.1. By comparing σb, the variance σ Select the color with the highest value.

(3) Using the average value of the color at that time as a threshold, mxm pixels are binarized depending on whether they are larger or smaller than the average value and used as area information.

(4) Find the average data for each of the two regions found in (3) and use them as two pieces of color information.

The above steps will be explained using FIGS. 9 and 10.

Figure 9 shows 4x4 image data 90 of each color (x, y, z).
~92, indicating the magnitude of the average value and variance value σ. Since the color with the maximum variance value σ is X, the average value of the data of
The image data 90 is binarized at 45.6°. As a result, area information indicated by 93 in FIG. 10 is obtained.

The values of XI + y + l z+ of C1 in FIG. 10 are the values of each color x, y, and
The average value of pixels of z is defined as X2 + y2+ in C2
Z2 indicates the average value of pixels of each color x, y, and z in the region 94.

In this way, the area information and the two color information C, C2 are obtained, and further compressed by vector quantization and LUT 70, resulting in the code (36 bits) shown in FIG.
is converted to

In this type of compression, a single image consisting of N x N pixels is cut into m x m small sections, and this m x m is converted into a fixed length code of 1 bit, so the compressed image data is as follows. The memory structure consists of bits in the depth direction of the address space. Therefore, image address information is retained, and random access performance such as image search and writing at arbitrary positions is better than normal variable code length compression methods such as MH, MMH, MR, MR, MMR, etc., and editing It can be said that it is ideal for work etc.

The reason for storing in two colors in this embodiment is to provide optimal expression for color line drawings, characters, etc. In other words, characters and line drawings are stored in two colors: the color of the character line drawing and the color of the background. It is expressed as Therefore, such a compression method reproduces resolution information well, and for a uniform color area, these two colors C, , C2 have close color values, and in the limit, C, = C2.

(This is when all pixels of mxm have the same value) Since this state can be regarded as one of the extreme states shown in FIG. 8, it can be processed as a general state using the code shown in FIG.

When using the compression method described above, if the restored image data is enlarged by a factor of 2 or 3 and printed out, a 4x4 (generally mxm) texture structure will appear and the image quality will deteriorate. It was found that this occurs. This is because the image information in 4x4 (generally mxm) is approximated with two colors,
This is because the 4x4 area is divided into two colors in the coloring picture, and the hue is not continuous with other adjacent 4x4 areas.

In other words, since the two colors are determined using only the information within the 4x4 pixel of interest, the colors become discontinuous with other adjacent 4x4 areas, and even within the 4x4 pixel of interest. This is thought to be because the two colors are represented in a coloring book style. As a result, when the composite image is enlarged, this 4×
4 boundaries and texture structure (moiré) become noticeable.

[Description of Decoding Circuit (FIG. 1)] FIG. 1 is a functional block diagram of the decoding circuit 103 of this embodiment.

For example, when the compressed encoded code shown in FIG. 8 is input, the restoring section 104 restores the color information c1-c2 to 24-bit color information and the area information 102 to 16-bit area information. 105 is a line memory that stores restored image data. 106 is a smoothing circuit that performs smoothing in units of a 3×3 pixel matrix in Example 1.

[Description of Restoration Unit (Fig. 2) Fig. 2 is a block diagram showing the configuration of the restoration unit 104.
Parts common to the figures are indicated by the same symbols.

20 to 22 are lookup tables (LUT) for restoration
Then, LUT 20 converts the 12-bit color information C1 into 24-bit CI (rl, g++ b,), and LUT 21 converts the 12-bit color information C1 into 24-bit CI (rl, g++ b,).
The bit color information C2 is converted to 24 bits C2 (rl, g2.
b2) respectively (however, here rl +
rl r gl + g2 +t)+, b2 are each 8-bit color component data). LtJT22 is 12
The bit area information 102 is restored to the original 4×4 (16 bits) area information. 24 is a parallel serial converter (
psc) converts the 16-bit area information into serial data in pixel order in a 4×4 matrix and outputs it.

23 is a selector, and serial data 2 from PSC24
7 and when the data 27 is 1゛, select the output 25 of LUT20, and when the data 27 is 0゛, select the output 25 of LUT20.
1 output 26 is selected and output. 105 is a line memory having a memory capacity for at least 8 lines x 3 colors when 4×4 compressed data is input as in this embodiment, and when the encoded code shown in FIG. 8 is input, Eight lines of image data are stored in the order of the overlapping pixels.

[Description of smoothing circuit (Figures 3 to 5)] Figures 3 and 5
The figure is a block diagram of a smoothing circuit that smoothes restored image data.

Eight lines of image data are read out from the line memory 105 and three lines are selected by the selector 30. This selector 30 selects a line in synchronization with a line synchronization signal (not shown) or the like so that three consecutive lines are always selected. Reference numeral 31 denotes a one-pixel delay circuit, which is configured to take out pixel data of 3×3 continuous small sections using axi signal take-out taps.

Figures 4(A) and 4(B) are diagrams showing examples of spatial filters for smoothing. Figure 4(A) is the one with weight placed in the center, and Figure 4(B) is the same. A filter that performs weighting is shown below. In general, placing more weight on the center requires more complicated hardware, but tends to result in better image quality.

FIG. 5 shows a configuration diagram when a convolution operation is performed using the spatial filter of FIG. 4(A).

50, 51, and 54 are adders, and 52.53 are coefficient multipliers.

The output of the coefficient unit 52 is 0.3x (a+c+g+i), the output of the coefficient unit 53 is 0.5X (b+d+f+h), and the adder 54 combines the output of the coefficient unit 52.53 with the center pixel (e
) is added. This means that the calculation shown by the following formula has been executed.

The normalization coefficient circuit 55 is a circuit that divides the output of the adder 54 by the number of pixels in the pixel matrix (nine in this embodiment). A smoothed output 56 is thus obtained, and this smoothing process is performed individually for each of the r, g, and b colors.

The above smoothing process has the following effects.

(1) What used to be a coloring picture based on two-color information within a 4x4 pixel now changes continuously.

(2) By making the matrix size for smoothing different from that for encoding, smoothing is performed across adjacent 4×4 pixels, eliminating boundaries between each 4×4 pixel.

Therefore, even if the image is enlarged, an image output in which the texture structure is not visible can be obtained.

Note that the larger the size of the spatial filter in this embodiment, the greater the smoothing effect and the better the effect, but on the other hand, the resolution decreases. It is preferable to select the size accordingly.

[Effects of the Invention] As described above, according to the present invention, there is an effect that image data encoded in such a manner that a high compression ratio can be obtained and at the same time efficient image editing can be decoded without deterioration in image quality.

[Brief explanation of the drawing]

Figure 1 is a functional block diagram of this embodiment, Figure 2 is a block diagram of the restoration section, Figure 3 is a circuit diagram for extracting 3x3 pixels, Figure 4 (A),
(B) is a diagram showing an example of a spatial filter. Figure 5 is a circuit diagram for inputting pixel data and performing addition and subtraction. Figures 6 (A) to (C) are image data converted into color information and area information. A conceptual diagram of the dividing process. Figure 7 is a diagram showing the configuration of the image data compression unit. Figure 8 is a diagram showing the code encoded by this embodiment. Figure 9 is the average value and variance of image data and pixels. FIG. 10 is a diagram showing an example of calculation results for area information and two-color information. In the figure, 20 to 22.70... lookup table (
LUT), 23.30...Selector, 24...Parallel-serial conversion circuit (PSC), 31...1 pixel delay circuit, 50,51.54...Adder, 52゜53
... Coefficient unit, 55 ... Normalization coefficient circuit, 60
．． 90.92... Image data, 102... Area information, 103... Decoding circuit, 104... Restoration unit, 10
5... Line memory, 106... Smoothing circuit. 7゜Figure 7

Claims (2)

[Claims] (1) An image obtained by dividing a predetermined pixel matrix of color image data into two pieces of color information at most, determining a color region based on the color information, compressing and encoding each of the color information and the color region. A data decoding method, comprising a restoring means for restoring the compressed color information and the color region to the original information, and a smoothing means for smoothing the restored information. method. (2) The decoding method according to claim 1, wherein the smoothing means smoothes with a matrix size different from the size of the encoded pixel matrix.