JPH03208469A - Picture data compression system - Google Patents

Abstract

PURPOSE:To attain valid data compression independently of the kind of picture by coding the data while studying the regularity such as a linearity or way of bent of a contour line in matching with an object picture. CONSTITUTION:An inputted picture is converted into a data with a 1st coding means 10 where the relation of connection of a contour line is traced and a rearrangement means 12 rearranges the data into the relation of connection along with the contour line and optimization is applied while studying the statistic property of the linearity or way of bent of the contour line with a 2nd coding means (universal coding) 16 and the relation of connection along the contour line is divided into a group of start point information only and a group of connection point information and end point information, they are rearranged respectively to be arranged and sets of information of different property are studied in the lump to enhance the effect of study thereby improving the compression rate.

Description

[Detailed Description of the Invention] [Summary] Regarding an image data compression method for compressing binary image data obtained by reading line scanning, the purpose is to reduce the amount of data by efficiently compressing the data, and to Since the connection relationship of the pixels that change between the two is MM configured, the compression efficiency deteriorates for images outside the expected range.

In contrast, in the present invention, as shown in FIG. 2(a)j, the input image is first converted into data traced by the contour line connection, and then the data is converted to data traced by the contour line connection, as shown in FIG. 2(b)a. The lines are rearranged into connection relationships along the contour lines, and the statistical properties such as straightness and curvature of the loop lines can be optimized while learning by unino-monkey encoding. ), the connection relationships along the contour spines are divided into a group containing only the start point information; a group containing the connection point information and the end point information; and they are rearranged so that they are grouped together, and information with a unique nature is learned together. By doing so, the learning effect is enhanced and the compression ratio is improved. Here, the start point is the position where the contour line first appears, the end point is the position where the contour line ends, and the connection point is the part where the contour lines connect between the start point and the end point. do.

Universal codes are originally an information-preserving data compression method that does not assume statistical properties of the information source in advance when compressing data, so it can be applied to various types of data (character codes, object codes, etc.) can do. In document images, there are similarities in the linearity and curvature of character lines.

Furthermore, in a halftone image, a huge number of changing points appear because the halftone dots are dispersed throughout the image, but the connections between the contour lines are similar due to the periodicity of the halftone dots and the same shape of the halftone dots. The redundancy caused by this similarity can be reduced by universal encoding and effective compression can be performed.

In the present invention, in order to find the connection relationship of contour lines, the MMR method is used to detect the mode of changing pixels as a preprocessing, and the connection relationship of changing pixels for each line is corrected to the connection relationship of the contour line itself, and this data is universally used. Encoding (Xi▼-Le
a+pel encoding) is applied, and the global properties of the straightness and curvature of the contour are expressed by the index of the universal code.

Here, the universal encoding method will be briefly explained. As a representative universal encoding method, zi
▼-There is a Lempel code (for details, for example, Munakata r
"Xi▼-Lempel's Data Compression Method", Information Processing, V
ol. 26. No. 1, Summer 985).

There are two types of Xiw-Lempel encoding methods: ■Universal type and ■Incremental decomposition type (Incremeajsl ptrsin)
Two algorithms have been proposed: g). here,
These two algorithms will be described.

[Universal Algorithm] Although this algorithm requires a large amount of calculation, it provides a high compression ratio. This is a method in which encoded data is divided into sequences (subsequences) of maximum length that match each other from an arbitrary position in a past data sequence, and encoded as a copy of the past sequence.

FIG. 3 shows the principle of a universal H▼-Lempel code encoder.

In FIG. 3, encoded input data is stored in the P buffer, and data to be encoded is input to the Q buffer. The Q-buffer sequence is searched through the P-buffer sequence to find a matching sub-sequence of maximum length in the P-buffer. The following set of information is then encoded to specify this maximum length subsequence in the P buffer.

Next, the encoded sequence in the Q buffer is transferred to the P buffer 7 to obtain new data. Thereafter, similar operations are repeated to decompose the data into subsequences and encode them.

Furthermore, as an improvement of the universal algorithm, LZ
There is an SS code (T.C. Bell, ' Bet
目 t OPM/L Text Compression
', IEEE Trsat. on Commu+,,
Vof. COM-34, No. 12, Dec.
(see 1986). In LZSS encoding, the start position of the maximum matching sequence in the P buffer, the matching length set, and the next symbol are distinguished by flags, and the one with the smaller amount of code is encoded.

[Incremental decomposition type algorithm] This algorithm has a lower compression rate than the universal type, but is known to be simple and easy to calculate.

In the incremental decomposition type 2iv-Lempel code, if the input symbol sequence is X=aabababaa * * ●, then the incremental decomposition into the component sequence x=Xo XI X2 . . . is performed as follows.

Let Xj be the longest string from which the rightmost symbol of the existing part is removed, and X=a1tabIIabaIIbIIaaII●●.

Therefore, Xo = λ (empty row), XI = XO a,X
2 =XI b, X3 =X2 a, X4 =
It can be decomposed as XO b,X5 =XI a, - - -. Each incrementally decomposed dynamic sequence is encoded as the following set using the existing component sequence.

That is, in the incremental decomposition type algorithm, for a coding pattern, among the subsequences decomposed in the past, the maximum length match is found and encoded as a copy of the subsequence decomposed in the past.

Furthermore, as an improvement of the incremental decomposition type algorithm, LZ
There is a W encoding (T. A. Welch, 'A
Technique for High-Perform
mance Data Compression',C
computer, John 1984). LzW
In encoding, the next symbol is incorporated into the next subsequence, so that it can be encoded using only the index.

That is, the global connectivity of the contours is represented by a fixed-length code as a subsequence of already encoded components or a component index, and the next symbol is represented by an MR code.

With such universal encoding, the linearity and curvature of characters, as well as the tendency of outlines forming halftone dots, can be captured and encoded as patterns, and effective compression can be achieved. In addition, the dictionary created by learning in universal coding refers to past sequences that have recently appeared in both the universal type P-buffer and the incremental decomposition type dictionary, so sequences containing components with different properties are learned individually for each component. Efficiency will improve if you do so.

[Example code] FIG. 4 shows the procedure of image data compression according to the present invention.

In FIG. 4, first step SL (hereinafter referred to as "step")
(omitted) is converted into a fixed-length mode code and a fixed-length RL code (run-length code) using an encoding method modified from MMR shown in FIG. These fixed length codes are in units of 1 byte, as shown in Figure 5(a).
are arranged like this. Incidentally, FIG. 5 is directed to the pixel arrangement shown in FIG. 6.

Next, the process proceeds to step 82, where the fixed length codes are rearranged line by line and contour by contour, resulting in the arrangement shown in FIG. 5(b).

Next, proceed to step 33, divide the group into a group containing only the starting point information, and a group containing the connecting point and ending point information, and summarize them in Figure 5 (
Arrange it like the array in C).

Finally, in step 84, the connection relationships having the arrangement shown in FIG. 5(c) are encoded using the universal encoding method.

Next, the processing from 81 to S4 in FIG. 4 will be explained in detail.

[Modified MMR encoding]: The modified MMR encoding method shown in FIG. 8 is different from the standard method shown in FIG. It is to make it a mode. That is, if the contour lines are connected, they will be represented in vertical mode. The contour line will follow the following mode transitions.

Horizontal mode → Vertical mode → Path mode (starting point)
(Connection point) (End point) Figure 6 shows the starting point, connection point, and
An example of the arrangement of end points is shown.

Fixed-length codes are assigned to each mode of changed pixels and run length (Run Lengtb) in modified MMR encoding, so that each code can be identified.
Expressed in bytes as shown in the table in the figure.

[Rearrangement and grouping along the contour line] FIG. 9 shows a processing flow for rearranging fixed length codes along the contour line used in the present invention.

FIG. 10 shows the sorting according to this processing flow.

Figure 10(a) shows how fixed-length codes obtained by the modified MMR method in Figure 8 are stored in memory, H: Horizontal mode (starting point) V: Vertical mode (connection point) P: Path Indicates the mode (end point).

Further, FIG. 10(b) shows the connection state of contour lines in an actual image (0 is indicated by an arrow).

In the modified MMR encoding method, connection of change pixels is determined line by line, so the change pixel that appears i-th in each line does not always belong to the same contour line.

The array B (1, k) in FIG. 10(c) is rearranged so that the k-th changed pixel in the array belongs to the same contour in each line. The digits of this array B are obtained by scanning the image of FIG. 10(a) from the upper line to the lower line,
Take the changing pixels in the order they appear. For example, it is clear from FIG. 10(b) that when scanning line No. 1, five changed pixels appear, so digits 1 to 5 are taken. Also line no. 3 scan, H2. , H2, and two new pixels appear, so add two digits to 1 to 7.
I take the.

Line No. Horizontal mode H2. , H2, when a new changed pixel appears, the order in which the changed pixels are arranged is shifted, but the order at this time is managed by the array CN shown in FIG. 10 (d).

That is, the array CN(m) is allocated so that the appearance order m of the changed pixel actually yields the digit k of the array B (1, k) in which the contour line to which the changed pixel belongs is stored. To explain specifically, line No. in FIG. 10(b),
3 scan, for digits 1 and 2, see Figure 10 (C
) matches digits 1 and 2. However, Fig. 10(b)
In digit 3.4, pixels newly changed due to horizontal modes H20 and H23 appear, so two contour lines are added,
Since the contour lines 1 to 5 have already been rearranged, the new contour line is placed in column 6.7 in FIG. 10(C). Therefore, the array CN is CN (P)=6 for P=3.4
．． 7, the two are associated with each other.

Further, when pass mode P is issued in line scanning, a signal empg indicating that the changed pixel has disappeared is stored in the first digit k of the line after pass mode P is issued for identification.

Next, a description will be given of the rearrangement process along the contour line of n lines of the fixed length code shown in FIG.

First, the number ml of changed pixels of the final line obtained when the previous n lines were rearranged in S1 and S2 is loaded into the variable m representing the number of contour lines. For the first n lines of the image, the initial value of ml is 1. Next, in S3, array CN contains 1 to m as the digit positions for storing the contour line in array B.
Set 1 to m1 in the 1 position, respectively. Subsequently, in steps 84 to S11, the fixed length codes of the changed pixels are read one by one and transferred to an appropriate position in array B.

If the corresponding position of array B, i.e. line 1, change pixel order i, B (1, CN(i) ) $::emp
If 811 determines that the ty signal is input,
After incrementing i in step 810, array B is adjusted again in step S11. If there is no empty signal in B(1.CN(i)), the process proceeds to S12 and subsequent steps to check the encoding mode of the fixed length code.

When the vertical mode ■ is determined in step 812, the process proceeds to S13, where the read fixed length code is stored in B(1, CN(i)).

Also, if the horizontal mode H is determined in 814, it means that a new pixel has appeared, so the process advances to s15 and the CN(I)
) is set to m+1, and the fixed length code of the new changed pixel is stored in S16. Also, in 815, the original C N (
The storage location of B digits from i) to CN(m) is set to the original CN.
Shift by one after (i+1).

Furthermore, if the encoding mode is pass mode P, the contour line will disappear, so proceed to S17 and convert B(1. CN(
i) ) and the subsequent lines of B (1, CN (i+1) ).

After repeating the above process, sorting n lines and storing the fixed length code in array B, proceeding from S6 to S19 to S31, the same CN (fixed length code from array B every n lines along D) By reading the code, a fixed length code is output along the contour line.

For horizontal mode when reading fixed length codes,
As shown in FIG. 11(a), line numbers and pixel numbers are collected in the memory area A of the starting point information so that it can be known in which line (Y) and in which pixel (X) it has appeared. Further, the vertical mode is stored as connection point information, and the path mode is collected and stored in memory area B as end point information.

After storing the fixed length codes of all contour lines, as for the starting point information, calculate the number N of contour lines, the relative value ΔY of the line number, and the relative value ΔX of the pixel number, as shown in FIG. 11(b). , read out in this order with a fixed length. Next, Fig. 11(a)
The connection point information and end point information stored in memory area B of are read out in sequence for each contour.

The information read out in this way is compressed using a universal encoding method and output.

Although encoding has been described in the above embodiments, the restoration (decoding) of image data can be accomplished by performing the procedure shown in FIG. 4 in reverse. That is, after decoding the fixed-length code along the contour line, the fixed-length code is rearranged line by line, and the modified MM
Decode using the R encoding method.

In addition, in the explanation of the embodiment of the present invention, when rearranging fixed-length codes, output is performed sequentially starting from the left contour line for each plurality of lines. It is also possible to give priority to the output as it is further to the left.

[Effect] As explained above, according to the present invention, the linearity of the contour line,
Since encoding is performed while learning the regularity of the degree of curvature according to the target image, effective data compression can be achieved regardless of the type of image.

Furthermore, since contour lines are grouped into starting point information, connecting point information, end point information, and different information representing the contour line, efficient learning of contour line rules is possible and a high compression rate can be obtained.

Furthermore, the regularity of the contour line is specified as a copy of the regularity of the encoded contour line, so even when the resolution increases, the regularity can be promoted as a group, and the amount of code is proportional to the resolution. There is no increase in the amount of data, and highly efficient encoding can be performed.

Furthermore, since the encoding tracks the contour line, there is no need to target the entire image, and it can be processed for each line using the MMR method, so there is no need for a large capacity image memory. High-speed processing can be performed by serial processing on a line-by-line basis.

[Brief explanation of drawings]

Fig. 1 is an explanatory diagram of the principle of the present invention; Fig. 2 is an explanatory diagram of the operation of the present invention; Fig. 3 is an explanatory diagram of universal encoding used in the present invention; Fig. 4 is an explanatory diagram of the image data compression procedure of the present invention. Figure; Figure 5 is an explanatory diagram of changes in the data arrangement according to the procedure of the present invention; Figure 6 is a diagram of the arrangement of the start point, connection point, and end point of image data; Figure 7 is the original MMR deformed encoding according to the present invention. An explanatory diagram of fixed-length code allocation of the method; FIG. 8 is a flowchart of encoding processing of the modified MMR method of the present invention; FIG. 9 is a flowchart of fixed-length code rearrangement processing along the outline of the present invention; FIG. 10 is an explanatory diagram of the sorting process in FIG. 9; FIG. 11 is an explanatory diagram of expression format conversion during universal encoding of the present invention; FIG. 12 is an explanatory example of the conventional MMR encoding method; FIG. 13 is an MMR code Explanation diagram of mode definition of conversion: Figure 14 is MMR
An explanatory diagram of fixed-length code allocation for encoding; FIG. 15 is a flowchart of encoding processing in the MMR method; FIG. 16 is an explanatory diagram of predictive encoding; FIG. 17 is a circuit configuration diagram of adaptive predictive encoding. In the figure, 10: first encoding means 12: sorting means 14: organizing means 16: second encoding means

Claims (3)

[Claims] (1) In an image data compression method that compresses binary image data obtained by scanning a reading line, a first step is performed to determine the connection relationship of changing pixels between adjacent scanning lines of the binary image data for each scanning line. an encoding means (10); a rearranging means (12) for rearranging the data encoded by the first encoding means into a connection relationship of changing pixels along the contour line of the image; and a rearranging means (12); Organizing means (12) that divides and summarizes the connection relationships of changing pixels sorted along the contour line into starting point pixel information, connected pixel information, and end point pixel information.
14); and second encoding means (16) for encoding the connection relationships organized by the organization means (14) as copies of connection relationships that have already been encoded. Additive data compression method. (2) The second encoding means (16) encodes the connection relationship to be encoded at the present time as information specifying a copy position and a copy length of the already encoded connection relationship. 2. The incremental data compression method according to claim 1. (3) The second encoding means (16) encodes the connection relationship to be encoded at the present time as information specified by the number of the subsequence when the already encoded connection relationship is divided into different subsequences. 2. The image data compression method according to claim 1, wherein the image data compression method comprises: