JPS5929020B2 - 2D block encoding method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION An object of the present invention is to improve data compression efficiency in two-dimensional binary image information such as facsimile signals by performing pattern encoding using fixed-sized two-dimensional blocks as encoding units. .

Conventionally, in order to transmit two-dimensional binary image information such as a facsimile, there are three methods: one method scans every 11 lines and converts it into a one-dimensional signal sequence and encodes it, the other method encodes 22 lines at once. There are many known examples of data compression, such as a method that encodes only the difference information from the previous line, a method that predicts the next pixel from 4 known pixel information, and encodes only the information when the prediction is incorrect. be.

However, there are problems such as limited compression efficiency and error propagation in reproduced images due to transmission errors. The present invention is capable of increasing data compression efficiency in view of the above-mentioned prior art, and details of the present invention will be explained below using a 4×4 block as an example.

FIG. 1 shows how original image information is divided into 4×4 basic blocks, which are coding units. The basic block encoding method utilizes the statistical properties of pattern occurrence frequency to assign codes with short code lengths to patterns with high frequency of occurrence, and codes with long code lengths to patterns with low frequency of occurrence. , specific examples of the encoding method are shown in FIGS. 2 and 3.
FIGS. 2a, b, and c are diagrams in which the 4×4 head-shaped patterns are divided into classes and arranged in descending order of frequency of occurrence. Note that the figures obtained by rotation are also the same pattern, and the respective pattern numbers are the numbers at the bottom right of each figure. As shown in Figure 3A and b, the encoding method for pattern numbers consists of a code corresponding to the class, then a code corresponding to each pattern within the class, and finally a code corresponding to the direction of rotation. shall be. FIG. 4 shows a diagram in which the encoding example described above is applied to FIG. 1. The 1st bit of the 1st row indicates the 0th class, the first 6 bits of the following 13 bits are the pattern following the 5th class, the next 5 bits are the 10th pattern in the class, The last two bits indicate that the direction remains as specified. The next 3 bits indicate that 3 of the 0th class will continue. A modification of the encoding method shown in FIG. 3 will be described. In the encoding method explained in Figure 3, a code with a code length of 1 is assigned to a pattern in which all 4x4 blocks are white, but the frequency of occurrence of this pattern is by far the highest compared to other patterns. Considering this, as shown in FIG. 5, a more efficient data compression method can be obtained by adding a code corresponding to the class only to this pattern followed by an encoded run length of the pattern. Next, an embodiment of the encoding method explained in FIG. 3 is shown in FIG.

Input pattern I given in 4X4 block.

-115 is a bit pattern X that is rotated in four different directions by the block rotation circuit 1 and has the minimum value thereof. -X
, 5 and bit D corresponding to the rotation direction. ，
, is output. On the other hand, the counter has bit pattern I. -1
, 5 are input, and the strobe signal ST is generated at the same time as
B counts the clock from clock generator 2.
It starts to pick up. Output A of counter 3. , 5 are input as address signals to the ROM 4 which stores bit patterns. The memory contents of ROM4 are as shown in Figure 7.
This is a bit pattern corresponding to the pattern in the figure. Address signal A. 16 output from ROM4 in response to ~5
Bit data Y. ~15 is the pattern X determined by the comparator 5. .about.15, and outputs the signal EQ only when they match. The encoding circuit 6 outputs the output A of the counter 3 at the time when the coincidence signal EQ is generated. ~5 and rotation direction information D. ，，
is encoded according to FIG. 3, and the signal sequence S is output together with the clock C. If the match signal EQ is not generated, the fixed pattern 111111 is followed by the original bit pattern I.
~15 is output as is. As described above, the present invention 2
A dimensional block encoded signal can be obtained. As described above, since the two-dimensional block encoding system of the present invention does not rely on differential data for redundancy removal, error propagation in reproduced images due to bit errors does not occur.

Also, in order to encode the two-dimensional block as a pattern,
Highly efficient data compression can be performed.

[Brief explanation of the drawing]

Figure 1 is a diagram showing how a two-dimensional image signal is divided into plotters, Figure 2 is a diagram showing the correspondence between 4x4 block patterns and pattern numbers, and Figure 3 is a diagram showing the correspondence between pattern numbers and codes. , FIG. 4 is a diagram showing the code sequence when the encoding examples shown in FIGS. 2 and 3 are applied to FIG. 1, and FIG.
A diagram showing an encoding method applying the encoding method shown in the figure, FIG. 6 is a diagram showing an embodiment of the block encoding method of the present invention,
FIG. 7 is a diagram showing the memory contents of the ROM 4 in FIG. 6. STB: strobe signal, S: signal series, C: clock.

Claims (1)

[Claims] 1. Original image information of a two-dimensional binary image signal such as a facsimile is divided into m×n two-dimensional blocks, and all representative m×n two-dimensional block patterns are divided in advance.
The patterns obtained by rotating are considered to be the same pattern,
Several patterns are grouped together to form a class, and for the representative pattern and the pattern obtained by rotating it, first, we correspond to each class that has the property that the code length is short for classes with a high frequency of occurrence. A code consisting of three parts is used: a code that identifies each pattern within the same class, and finally a code that corresponds to the block rotation direction if a different pattern is produced by rotation. A two-dimensional block encoding method characterized in that patterns other than the above-mentioned representative patterns are comprised of two types of parts: a code that identifies the other class, and a code that uses the m×n bit pattern as it is. . 2. The two-dimensional block encoding method according to claim 1, wherein each block is a 4×4 block.