JPS6080370A - Picture information compressor - Google Patents

Abstract

PURPOSE:To obtain a variable length data with high speed efficiently and to attain a high speed editing processing by processing a picture digital data transmitted in the unit of words in the lump in bit parallel. CONSTITUTION:An original picture data to be comressed is stored continuously in the main storage of a host computer 1 or equivalent storage device in the order of scanning at each line. The host computer 1 transfers the original picture data in the unit of words to an image compressor 2 by an input/output channel, the transferred data is converted into an MH code or an MR code according to the attribute of a picture by an encoding means 4, and a variable length code data 5, a code length 6 and control information 7 are given to editing means 8 of the next stage. The editing means 8 edits the code data in a width possible to be transmitted to the host computer 1 based on the information above and transfers the result to the host computer 1 as a series of code data 9.

Description

[Detailed Description of the Invention] Technical field to which the invention pertains The invention relates to an image information compression device VC.5%
The present invention relates to an image information compression device that uses run-length coding and vertical reference coating to compress binary digital data obtained by fully scanning a pfT foot image.

The conventional run-length coding method is a method in which a continuous run of black or white pixels is represented by a code word of variable length corresponding to the run length, and a short code word is assigned to a code word that appears frequently.

In addition, the vertical reference coating method allows the color change position in the current scan line, ie from black to white.

This method assigns a code word depending on how far the position where the color changes from white to black is relative to the color change position in the previous scanning line.

The former run-length coding method is called ``6-motivated Huffman coding''.

The latter vertical reference coding method is a 6-mode, 7-eyed
This has been adopted by the CCITT (International Telegraph and Telephone Consultative Committee) as ``Lead Coding.'' Compression technology using these coating methods has conventionally been mainly realized in the same line systems such as facsimiles, and data Since the transfer is serial, it is sufficient to perform the conversion process serially one bit at a time.However, recently there has been a tendency to use these coating methods for image processing by computers, and the entire data transfer during -t is It has become necessary to do this in parallel.

For example, the original image digital data stored in word units in the memory unit is transferred to the compression device in parallel via the ta-speed data bus, and the compressed code data is all transferred in parallel from the compression device to the memory unit. There is an operation of transferring data via the same data bus. In this case, if a conventional serial conversion tube is used as a compression device, the conversion process will count all of the parallel data one bit at a time, check the attributes of that data, code it according to the attributes, and then output the code one bit at a time. This is done by editing the data into fixed length data that matches the data bus width using a shift register or similar tube.This method can be implemented with relatively simple hardware, but it requires conversion in order to process data bit by bit. The drawbacks are that it takes a long time to process and is not suitable for computer image processing.

Purpose of the Invention The purpose of the invention is to process image digital data sent in word units in parallel pits to obtain variable length code data at high speed and efficiency, and to convert this variable length code data into its code length information. It is an object of the present invention to provide an image information compression device which eliminates all of the above-mentioned drawbacks by editing data into predetermined fixed length data using sound.

Structure of the Invention An image information compression device according to the invention converts digital data obtained by fully scanning a predetermined image into attribute VC of the image.
An encoding means for generating variable length code data and its code length information using the Nth compression method, and a plurality of variable length code data successively generated by the encoding means using the code length information. editing means that operates to add at least a part of the following data to the end of the preceding data among the successive variable length code data in order to output them as a plurality of mutually divided fixed length data; It is all inclusive.

More specifically, image data is processed in batches for several pixels, all color change positions in the processed pixels are determined, the attributes are analyzed from the borrowed color change position information, and the attribute is matched. Variable-length code data, its code length information, an encoding means that generates all the variable-length code data, and variable-length code data and code length information that are continuously generated by the encoding means, and these continuous variable-length code data are encoded into a plurality of fixed pieces. In order to output long data, at least a part of the following variable length code data is added to the end of one variable length code data to form one fixed length data, and the remaining data part is added to the following variable length code data. It is stored so that it is added immediately before the code data, and if it does not reach the fixed length data even if all of the variable length code data is added to the end of the previous data, it is not output as fixed length data and the next and editing means for storing one copy of the following data to be added.

Examples of the invention The present invention will be explained below with reference to the drawings.

FIG. 1 is a schematic block diagram illustrating the principle of non-invention. In FIG. Continuously stored in FF3. The host computer 1 transfers the original image data in units of image compression devices g2 (rr-words) at a speed of h using channels as used in general input/output operations. In this embodiment, the word length is assumed to be 16 bits. The transferred original image f-3 is.

2) The encoder 4 converts the image into a modified Huffman (M tricode) or modified lead (NFL) code according to the attributes of the image, and converts the variable length code data 5, code length 6, and control information 7 into the following code. Pass it to the editing means 8. The editing means 8 is the encoding means 4.
All the variable length code data 5 and code length 6 continuously sent from the host computer 1 are used to cut and paste the code data so that it has the width (16 bits in the actual column) to be transferred to the host computer 1. The code data 9 edited into a fixed length by the editing means 8 is transferred from the image compression device 2 to the host computer 1 via the channel in the same manner as the transfer of the original image data, and is stored in an appropriate storage device.

Next, details of each configuration will be explained.

FIG. 2 shows a block diagram of the encoding means 3 of FIG. 1. *FIFO (First In First)
Jut) 10 is simply for buffering the transferred original image data 3. The delay circuit 11 delays all data for one line in order to perform MR cocoing, and in the embodiment, for simplicity, it is assumed to be an integral multiple of the total word length of one line. Therefore, the delay is performed word by word by the number of words included in one line.

The original image data 12 and 13 for one word of the previous line delayed by the current line and the delay circuit 6 undergoes a pixel color transition from white to black or from black to white by the color change position detection means 14 and 15. Recognize what bit of VC the position corresponds to from the beginning of the line, and store the position information 1
Output 6.17. Naturally, it is possible that multiple color change positions are recognized in a VC in one line, and in that case, all color change position information in the word is output in order from the 6th position closest to the beginning of the line. .

FIG. 3 shows an example of a detailed circuit configuration of the color change position detection means 14 and 15 shown in FIG. ■1 is one word (16
represents the original image data (bit).

■! It is detected whether or not there has been a color change by performing an exclusive OR on each bit with the m previous bit, and the result is stored in the buffer 22. That is, 1" is output for 1a1tK where there is a color change, and "0" is output at a position where there is no color change. If the word data is at the beginning of the line, white (20) is output to the virtual pixel 21. ) is not the beginning, the color of the last bit of the previous word data is defined.

The color change position information stored in the buffer 22 is output in order from the position closest to the beginning of the line by the following method.

First, the output V2 of the buffer 22 is masked by the mask circuit section 27 for all bit positions where ++11+ is set in the output V5 of the buffer 25.

The masked state V3 is stored in the buffer 23 by the 5th order ascending priority circuit section 28, which displays only the 1" position closest to the beginning of the line among the bits in 3 where IT II+ is set. .In the beginning, Ba.

All 10'' bits are contained in the FF 25, so the mask circuit section 27 does not mask any bits of v2.

The output ■4 of the buffer 23 is set to the state V by the OR circuit 24.
5 and the result is stored in the buffer 25. That is, in the next cycle, state V5 displays 1" at the position closest to the beginning in state 2, so the mask circuit section 27 masks all the positions, and the ascending priority circuit section 28 displays 1" in state V2. , the second closest “1” to the beginning of the line
Full output. The above repetition is monitored and controlled by the comparator circuit 26. Comparison circuit 26Fi status v4 is all 0
++, and if all are not 0'', that is, if 1'' is set in any bit position, information 29 indicating whether that position is the pJth bit counting from the beginning of the line is sent. After that, the buffer 23.OR circuit 24
．． All clocks are sent to the buffer 25, and the next color change position is detected. If the total comparator circuit 26 detects that the state MV+ is all 0'', this means that the word data no longer has a color change position to be detected, so the control to request the next word data is executed. Generate 30 areas and clear 25 buffers in circulation.

Next, an example in which the color change position of the original image data is output by the color change position detecting means will be explained while following the operations in each state (1) to (V5). First, the state Vt of one word of the original image data is temporarily added as follows.

Vl:0O11100001100000 Virtual image search 2
Assuming that 1 contains N'OI+, the state ■
2 becomes V2:0010010001010000. The virtual double system 21 stores ``0'', which is the same as the last bit of ■1. Since the buffer 25 initially contains all 0'', the state MAVs is Vl: 0O10010001010000, the same as V2. Next, in state ■, the ascending priority circuit unit 28 controls the top vcH (
The 3rd bit is detected as '1' which is close to '1', and the state V
4 becomes as follows.

V4: ooiooooooooooooo.

The comparing circuit 26 recognizes that V4 is set to "1" and outputs "3" as one color change position information.Furthermore, in order to detect all the next color change positions, V4 is ORed with V5 by the circuit 24. It is ORed and stored in buffer 25. The new output V,l of buffer 25 is the same as v4.

V,':0010000000000000Vs' masks the third bit 11'' of the signal V2 by the mask circuit section 27.

That is, the following various new states Vs''e of Vl occur.

Vs': 0000010001010000■3' is a new state V of V4 such as 5th order by the ascending priority circuit section 28
4”fr occurs.

V4': OOOO010000000000■4'
The comparison circuit 26 recognizes that 1" is set and outputs 1 color change information "6".

Further, V,/ is logically summed with V,' by the OR circuit 24VC, and a new state V, //T of 5' is generated.

V,'': Perform all of the following operations 0010010000000000 times.

Vl:0000000001010000■4":0
000000001000000 Color change position ``10'' Output V5'': 0010010001000000V3
”': 0000000000010000v,”
:0000000000010000 Color change position ″12
"Output V,"": 0010010001010000v3
”” :0O00000000000000v 4””
: o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o o
o o o o o.

Here, the comparison circuit 26 recognizes that V4//I/ are all 0'', clears the buffer 25, and requests the next word data. Then, the color change position detection operation of the word data is completed.

Return to Figure 2 again. Color change position detection means 14 and 15
The color change position information 16 of the previous line and the color change position information 17 of the current line generated by
is input. The attribute detection means 18 analyzes all attributes of the image according to whether the coating method of the current line is the M)T method or the MR method. In general, the vertical reference coating method refers to all previous lines and converts the current line to 'kt-3', so if an error occurs in one line, it affects all subsequent (υ) lines. Therefore, the run length is changed every few lines.・A method is used in which coding is inserted to prevent the spread of erroneous requests.In the MR method, a line by MH cocoing is inserted for every line. 24 is often used.The K value is set before compression.The attribute detection means 18 detects whether the current line is MH
If specified by the method, color change position information 16 of the previous line
is completely ignored, and the run length and its wrapper are detected from this color change position information 17 of the current line. Also, if the current line is instructed to the MR method, from information 16 and 17,
Analyze all their positional relationships, horizontal mode (H mode),
Either vertical mode (V mode) or path mode (P mode) is fully detected, and each mode is further classified into attributes in detail (see CCfTT Recommendation T, 4).

FIG. 4 shows the claw of the attribute detection means 18 of FIG. aO-al*a2 represents the color change position on the current line, and bl and b represent the color change position on the previous line. "C line" and "R line" respectively represent the operation of reading the color change position of the current line and the operation of taking the color change position d of the previous line.Po and PR represent the colors at the positions indicated by aQ and bl, respectively. A detailed explanation of the flow will be omitted.

FIG. 5 shows an example of the attribute information outputted by the attribute detection means 18. The MH/MR selection bit 31 completely identifies whether the MH method or the MR method is selected for the current line. P mode 32. H mode 33. ■Mode 34
Each identification pit is MR.

is valid only when selected, and which of the three modes is selected is determined at the time of attribute detection. ■The mode attribute identification field 35 has a 3-bit area, and is used as the relative shift amount of the current line color change position with respect to the previous line color change position calculated by block 71 in FIG. 4 when the mode is added. Displays one of -3 to +3. The RL field 37 and the color bit 36 represent the run length and its envelope f' when H mode or MH mode is selected in MI (method). That is, block 7 in Figure @4
2.71) Alternatively, the results and colors calculated in block 74 are displayed. The EOL bit 38 indicates whether the displayed attribute information is the end of the current line.When displaying the entire run length, it is the number of bits that can display the maximum length of the run. More than one.

By dividing the run length into multiples of 64 and their remainders and displaying them, the number of display bits can be shortened.

That is, a 6-bit area that displays several numbers from 0 to 63,
It can be displayed with a total of 7 bits, 1 bit identifying whether it is a make-up (quotient divided by 64) or a terminator (remainder).

The attribute information 19 output from the attribute detection means 18 is converted into CCITT Recommendation T, 4 by the following code conversion table 20.
It is converted into a variable length code as shown in , and at the same time, the code length is also calculated and output.

J: Sea urchin MI-1 code and MR evident from the code table
In the case of a code, the code length is at most 13 bits,
Therefore, it is sufficient for the fiJ variable length code data and the code length to have an area of 13 bits and 4 bits for iL, respectively. It is preferable that the code data format has a width of 16 bits, is left-justified, and the remaining 9 areas less than 16 bits are all filled with 0''.

FIG. 6 shows the flow of the editing means 8 of FIG. Ml and ni of the block 41 represent the i-th variable length code data and its code length, respectively. m in the block 42 represents the bit position to which the code data has been filled by the previous editing, and the bit position n that will be filled when the new code data Mi k is continued. In calculation/judgment 43, it is determined whether or not n exceeds 16 bits, thereby determining whether data for one word is satisfied. If it is satisfied, the shift operation of 44.45 is performed, and the logical sum of MO and Mi'' is outputted by 46.
Mt' and Mt'' are respectively divided into Mi mira half m bits and second half (16-m) bits, and Mj' is M
The empty area of O is fully complemented and output, and MO is updated to M j '' for the next editing operation. If MO does not reach one word even after continuing Mi, shift operation 47 is performed. The logical sum of MO and Mn2 is used as MO again to perform the first editing operation. If it is determined in judgment 48 that the page is over, output MO'< less than one word and end the editing operation. do.

FIG. 7 shows a detailed block diagram of the editing means 8.

The manually entered code length information is stored in the Mi buffer 51, and the value of the m buffer 52 is used.

Right shifter 53. and left shifter 54 by m bits and (16-m) bits, respectively. Here, it is desirable that the shifter be capable of shifting -i by a given number of shifts, and each result is sent to Ml' buffer 5.
5 and Mi" buffer 56. On the other hand, the code length information is first stored in the ni buffer 57, and the code length information is stored in the adder 58.
1 m bar.

It is added to the value of the buffer 52 and stored in the n buffer 59. 59 can count up to 15 numbers, and when a number greater than that is input, an overflow boost 60 is generated. This overflow signal acts on the ld multiflexor 6]C to complete the entire editing procedure. That is, when there is no overflow,
Output of Mi' buffer 55 and Mo buffer 63 = i(J
The result of the logical sum performed by the R circuit 61 is stored in the MO buffer 63.

In the event of an overflow, the result of the OR circuit 61 is sent to the FIFO 64 as data to be output to the host computer:
h, the output of the ``Mi'' buffer 56 is stored in the MO buffer 63. Furthermore, although not shown in the figure, control information indicating the end of the page is stored.

Zero or more edit operations cause the contents of the Mo buffer to be sent to its '24 F'
The fixed length data buffered in the FO is sequentially output to the host computer.

As described above, according to the present invention, each bit of original image data and variable-length code data is processed in parallel in batches, thereby enabling high-speed encoding and editing, which has the advantage of shortening processing time.

[Brief explanation of drawings]

1 is a schematic block diagram showing the entire principle of the present invention, FIG. 2 is a block diagram of the encoding means shown in FIG. 1, FIG. 3 is a diagram showing a complete circuit example of the color change position detection means shown in FIG. 2, and FIG. The figure is the second
FIG. 5 is a flowchart showing the processing procedure of the attribute detecting means shown in FIG. 2; FIG. 5 is a diagram showing an example of the attribute information output by the attribute detecting means of FIG. 2; FIG. 6 is a flowchart showing the processing procedure of the concentration means of FIG. 1, and FIG. 7 is a concrete block diagram of the editing means. Explanation of symbols of main parts 1...Host computer, 2...Image compression device, 4...Encoding means, 8...
...A means of bias. No. 74 Kokuso 6 Figure

Claims (1)

[Claims] Directly at an image information compression device that converts digital takes obtained by scanning a predetermined image into variable-length code data using a compression method layered according to the attributes of the image, a predetermined number of the images among the digital takes are converted into variable length code data. encoding means for collectively processing each data group corresponding to each pixel to generate variable-length code data representing the character and its code length information; In order to output the variable-length codetakes of ``b'' as a plurality of mutually divided fixed-length takes using the code length information, merge them into the end of the preceding data among these successive variable-length codetakes. and editing means operable to add all of at least one part of the take.