JPH0795412A - Image data compression method - Google Patents

Abstract

PURPOSE:To use the visual characteristic of human eyes to obtain a satisfactorily compressed picture with a smaller number of bits by limiting the change range of variable density to a minute block unit like a prescribed number of picture elements to quantize each picture element. CONSTITUTION:Data Y is divided into blocks B by a block dividing circuit 22, and a maximum value MAX and a minimum value MIN of data Y in each block B are taken out by a detecting circuit 23. Data Y from the circuit 22 is delayed and compensated by a delay circuit 24 and is divided into subblocks SB by a block dividing circuit 25, and a maximum value SMAX and a minimum value SMIN of data Y in each subblock SB are taken out by a detecting circuit 26. A D-range encoding circuit 28 quantizes SMAX and SMIN in the range determined by MAX and MIN and encodes these quantized values by degeneracy to a very small number of codes. A D-range decoding circuit estimates and outputs a maximum value SMAX' and a minimum value SMIN' of the subblock SB in accordance with the value of codes. A quantizing circuit 30 quantizes data Y of the subblock SB inputted from a delay circuit 27 within the range determined by SMAX' and SMIN'.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of compressing image data.

[0002]

2. Description of the Related Art In computer graphics, a color display capability of 256 colors per pixel is generally sufficient. In other words, it is sufficient to allocate 8 bits to 1 pixel.

However, recently, there is a demand for processing and displaying a natural image, for example, an image captured by a video camera, and in this case, the color display capability of 256 colors per pixel is insufficient.

However, since the computer graphics has a display capacity of about 2000 pixels × 2000 pixels, if the color display capacity per pixel is increased, the size of the display memory becomes too large. For example, if 8 bits are assigned to each of R, G, and B components in one pixel, 24 bits are required for each pixel, so that 24 bits x 2000 pixels x 2000 pixels = 96000 as a whole.
The size becomes 000 bits ≈ 11.4 megabytes, which takes a lot of time when transmitting the image data, or even if a hard disk device is used, only a small number of images can be stored. In any case, it becomes less practical.

Therefore, there is a compression method called the ADRC method as a method for compressing image data without degrading the image quality, particularly the resolution.

FIG. 3 shows an example of an encoder circuit for compressing image data by the ADRC method, in which the grayscale image signal Y is
One sample (one pixel) is 8 in the A / D converter 1.
It is quantized into bit digital data Y, and this data Y is divided into one block B [i] every 16 pixels (= 4 lines × 4 pixels) in the block division circuit 2 and further in the detection circuit 3. For each block, the maximum value MAX [i] and the minimum value MIN [i] of the data Y in that block are extracted.

Then, in the arithmetic circuit 4, 8-bit data D [i] indicating the dynamic range D [i] is extracted from these data MAX [i] and MIN [i]. Further, the data Y from the division circuit 2 is subjected to delay compensation for the detection circuit 3 in the delay circuit 5, and then the minimum value MIN is obtained from the delay-compensated data Y and the data MIN [i] in the subtraction circuit 6. The data (Y-MIN [i]) in which [i] and the following are cut off is extracted,
This is supplied to the adaptive encoder 7, and the data D is supplied to the encoder 7 so that, for each block B [i], each data Y has a level within the section D from the 8-bit data Y indicating an absolute level. It is requantized into 4-bit gradation data DATA based on MIN [i].

Therefore, for data Y of one block B [i], gradation data DATA of 4 bits × 16 samples and minimum value data MIN [i] of 8 bits MIN [i]
Bit dynamic range data D [i] is obtained, and these data are transmitted or stored.

FIG. 4 shows an example of a decoder circuit for obtaining the original grayscale image signal Y from the data DATA, which is the same as the original 8-bit data after the processing reverse to that of the adaptive encoder 7 and the subtraction circuit 6 is performed. Data Y of the D / A converter 14 and the data Y of the D / A converter 14
The original grayscale image signal Y is taken out even if it is supplied to.

Reference: Journal of the Institute of Electronics, Information and Communication Engineers, April 1987
Month issue pages 720-726, 736-741

[0011]

In order to improve the compression rate by the ADRC method, it is possible to increase the block size and reduce the number of quantization bits of the grayscale data DATA. However, as in the conventional example, the gradation data DA
When the number of quantization bits of TA is 4, the compression rate cannot be more than double, no matter how large the block size is. Therefore, in order to double the compression rate, it is indispensable to reduce the number of allocated bits of the gradation data DATA. The reduction of the number of quantization bits causes a quantization error. When encoding a block including contours, its dynamic range naturally becomes large, and the quantization resolution becomes coarse with a small bit allocation. Therefore, there is a problem that a block including a contour has a large quantization error of the entire block, and quantization noise is more noticeable than that of a block including no contour, resulting in an unnatural image.

The object of the present invention is to concentrate the quantization error only in the very vicinity of the contour, and by utilizing the visual characteristics of the human eye, image data in which a good compressed image can be obtained with a smaller number of bits. The purpose is to obtain a compression method.

[0013]

An image data compression method according to the present invention comprises means for dividing an image into minute blocks B, means for obtaining the maximum value MAX and the minimum value MIN of the minute blocks B, and the minute blocks B. Means for dividing into smaller micro blocks SB and maximum value S of the micro blocks SB
Means for obtaining MAX and minimum value SMIN, and maximum value SMA
Means for quantizing X and the minimum value SMIN within a range defined by the maximum value MAX and the minimum value MIN, degenerating the combination of the quantized values into a code of a very small number, and decoding the code, and decoding the maximum value. It is characterized by comprising means for obtaining SMAX 'and the minimum value SMIN', and means for quantizing the light and shade of the minute block SB within the range determined by the maximum value SMAX 'and the minimum value SMIN'.

[0014]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail below with reference to the drawings showing the embodiments thereof. FIG. 1 is a block diagram showing a configuration example of an image data compression circuit for carrying out the present invention.

The grayscale image signal Y is transferred to the A / D converter 21.
At, one pixel is quantized into 8-bit digital data Y. This data Y is divided by the block division circuit 22 into one block B for every 64 pixels (8 lines × 8 pixels), and further, in the first detection circuit 23, the data in the block B is divided for each block B. 8-bit data MAX and MIN indicating the maximum value MAX and the minimum value MIN of Y are taken out. Further, the data Y from the block division circuit 22 is subjected to delay compensation for the detection circuit 23 in the first delay circuit 24, and then the second
2, the data Y of the block B is divided into 16 sub-blocks SB [i] (2 lines × 2 pixels), and the second detection circuit 26 sub-block SB [i]. Maximum value SMAX of data Y for each i]
SMAX indicating [i] and minimum value SMIN [i]
[I] and SMIN [i] are taken out. In addition, the data Y from the block division circuit 25 is transferred to the second delay circuit 27.
In, the delay compensation for the detection circuit 26 is performed.

The D-range coding circuit 28 uses MAX and MI.
The threshold value that divides the range defined by N into four equal parts is th0, th
1 and th2, SMAX [i] and SMIN [i]
Depending on the value of CO, CO is divided into 8 cases as follows.
Output DE [i].

1) MIN≤SMIN [i] and SMAX
When [i] <th0 CODE [i] = 0 2) th0 ≦ SMIN [i] and SMAX [i] <th
When 1 is CODE [i] = 1 3) th1 ≦ SMIN [i] and SMAX [i] <th
When 2 is CODE [i] = 2 4) th2 ≦ SMIN [i] and SMAX [i] ≦ MA
In the case of X, CODE [i] = 3 5) MIN ≦ SMIN [i] and SMAX [i] <th
When 1 CODE [i] = 4 6) th0 ≦ SMIN [i] and SMAX [i] <th
When 2 CODE [i] = 57 7) th1 ≦ SMIN [i] and SMAX [i] ≦ MA
When X is CODE [i] = 68 8) MIN ≦ SMIN [i] and SMAX [i] ≦ MA
In the case of X, CODE [i] = 7 In the above example, the D range coding circuit 28 uses MAX and M
Although the range determined by IN is divided into four equal parts and the codes are assigned, the present invention does not necessarily have to be divided into four equal parts, and the number of cases need not be eight.

On the other hand, the D range decoding circuit 29 determines the maximum value S of the block SB according to the value of the code CODE [i].
MAX 'and the minimum value SMIN' are estimated and output as follows.

1) When CODE [i] = 0, SMAX '[i] = th0, SMIN' [i] = MIN 2) When CODE [i] = 1, SMAX '[i] = th1, SMIN' [I] = th0 3) When CODE [i] = 2, SMAX ′ [i] = th2, SMIN ′ [i] = th1 4) When CODE [i] = 3, SMAX ′ [i] = MAX, SMIN '[i] = th2 5) When CODE [i] = 4, SMAX' [i] = th1, SMIN '[i] = MIN 6) When CODE [i] = 5, SMAX' [i] = th2, SMIN '[i] = th0 7) When CODE [i] = 6, SMAX' [i] = MAX, SMIN '[i] = th1 8) When CODE [i] = 7, SMAX' [i ] = MAX, SMIN '[i] = MIN Obtained as described above MAX '[i] with the SMI
N '[i] represents the range of change of the data Y in the block SB [i]. The quantization circuit 30 uses SMAX '[i]
And SMIN ′ [i] within a range, each data Y of the block SB [i] input from the delay circuit 27 is, for example, 2
Quantize with bits, and use data DATA as the quantized representative value
[I] [j] (j represents pixel data in the block SB, j = 0 to 3) is output. As described above, the pixel data of the block B is transferred to the blocks SB [0] to S [S].
Sequentially encode up to B [15].

Therefore, with respect to the data Y of the block B,
2 × 4 × 16 gradation data DATA, 3 × 16 code data CODE [16], and 8-bit maximum value data M
AX and 8-bit minimum value data MIN are obtained, and these data are transmitted or stored.

Next, an embodiment of the expansion circuit according to the present invention will be described in detail. FIG. 5 is a block diagram showing a configuration example of an image data expansion circuit for carrying out the present invention. First, MA
X, MIN, and CODE [i] are input and supplied to the D range decoding circuit 31. The D range decoding circuit 31 is similar to the D range decoding circuit 29 in that SMAX ′ [i] and SMI
Output N ′ [i] [j].

The dequantization circuit 32 uses SMAX '[i],
Data DATA [i] [j] at the same time as SMIN '[i]
Enter. The inverse quantization circuit 32 uses SMAX '[i] and S
The quantized value corresponding to the data DATA [i] [j] is selected from the quantized values within the range defined by MIN ′ [i], and is set as the data Y of each pixel. The data Y is arranged in the block decomposition circuit 33 in the order of the original time axis by one sample, and the original grayscale image signal Y is taken out when the data Y is supplied to the D / A converter 34.

[0023]

As described above, according to the present invention, the variation range of the light and shade is limited to a small block unit of about 2 lines × 2 pixels and each pixel is quantized, so that only the outline portion is roughly quantized. Since the flat part near the contour is finely quantized, it is possible to perform quantization that matches the visual characteristics of the human eye, and the image can be compressed at a compression ratio of 2 times or more without degrading the image quality. It has the effect of compression.

[Brief description of drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of an image data compression circuit of the present invention.

FIG. 2 is a block division circuit 2 according to an embodiment of the present invention.
The figure which shows the block division | segmentation method of 5.

FIG. 3 is a block diagram showing the configuration of an embodiment of an image data expansion circuit of the present invention.

FIG. 4 is a block diagram showing a configuration of a conventional image data compression circuit.

FIG. 5 is a block diagram showing the configuration of a conventional image data expansion circuit.

[Explanation of symbols]

 1 A / D converter 2 Block division circuit 3 Detection circuit 4 Subtractor 5 Delay circuit 6 Subtractor 7 Adaptive encoder 11 Adaptive decoder 12 Adder 13 D / A converter 21 A / D converter 22 Block division circuit 23 Detection circuit 24 Delay circuit 25 block division circuit 26 detection circuit 27 delay circuit 28 D range encoding circuit 29 D range decoding circuit 30 quantization circuit 31 D range decoding circuit 32 inverse quantization circuit 33 block decomposition circuit 34 D / A converter

Claims (1)

[Claims] 1. A means for dividing an image into minute blocks B,
A means for obtaining the maximum value MAX and the minimum value MIN of the minute block B, and the minute block B having a smaller minute block S.
Means for dividing into B and maximum value SMA of minute block SB
A means for obtaining X and the minimum value SMIN, and a maximum value SMAX and a minimum value SMIN are quantized within a range determined by the maximum value MAX and the minimum value MIN, and the combination of the quantized values is degenerated into a very small number of codes and encoded. Means, a means for decoding the code to obtain a maximum value SMAX ′ and a minimum value SMIN ′, and means for quantizing the light and shade of the minute block SB within a range determined by the maximum value SMAX ′ and the minimum value SMIN ′. An image data compression method characterized by the above.