JPS6051826B2 - encoding device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to an encoding device that efficiently encodes a symbol sequence from an information source that has two symbols and in which the frequencies of appearance of the symbols are greatly different.

Now, if the appearance probability of symbol 1 is very small (<112) in a binary information source consisting of symbol 1 and symbol o, then
The probability of consecutive symbols O increases.

Therefore, by encoding the number of consecutive 0 symbols between symbols according to a code format that has a bit length allocation according to the frequency, the number of consecutive 0 symbols is It is possible to reduce the number of bits after encoding (compression). This method is called run-length encoding. That is, n of symbol 0 of binary information source
(n is an integer value on the 0 view) and the following symbol Jbol 1 are set as a new symbol rn (called a run-length symbol), and the above symbol is added to the symbol sequence of the output of the original binary information source. rn (the binary symbol sequence corresponding to each is called a run), and this sequence of run-length symbols is encoded by regarding it as an output from a new information source (called a run-length information source). be. but,
An encoding device using such a run-length encoding method can be configured only for binary information sources in which the number of run-length symbols that appear when considering the run-length information source is limited; Not applicable to sources.

Therefore, a sequence of symbols O (7) M bits (M is an integer value, M≧2) of the binary information source is set as a new symbol SM, and symbol 0 is i bits (i is an integer value, 0>1<M−1 ) followed by one bit of symbol 1 as a new symbol Si
Considering an information source limited to M+1 symbols, 2
A more practical method is to convert the original symbol sequence into the M+1 new symbol sequence and encode it. 2m order (
(m is a positive integer) A local code is one example of this, and its code format is shown in FIG.

Here, the value of M is set to a power of 2 (2m). And the symbol S above. -SM-, is assigned an m+1-bit codeword, and SM is assigned a 1-bit codeword. However, the above symbol S in Figure 1. - In the code word corresponding to SM-1, the m bits following the first 1 are the number of bits 1 of the symbol O before the symbol 1 of the binary symbol sequence, which is 2.
It is expressed in base code and arranged in order from the most significant digit.
Here, the symbol S is considered when encoding using the local code shown in FIG.

- An information source consisting of SM is named a local information source. FIG. 2 is an example in which M is set to 4 (m=2).

FIG. 3 shows an example of encoding using the code format shown in FIG.

In FIG. 3, a represents a symbol sequence of a binary information source, b represents a symbol sequence of a run-length information source, c represents a symbol sequence of a local information source, and d represents a local code sequence. Here we indicate the range of the symbol sequence of the original binary source that corresponds to the symbol of the curly bracket, run-length source, or local source. Arrows indicate codes corresponding to symbols of local information sources. However, according to such an encoding method, if the last block of a binary scene symbol sequence ends with symbol 0, there is a possibility that less than M symbols O remain. For example, in the case of the example shown in FIG. 3, the last 3-bit symbol O cannot be encoded. Conventionally, in order to encode this, a method has been used in which symbol 1 is added to the end of the symbol sequence. Therefore, in the case of Figure 3, the end of the symbol sequence is assumed to be 6'000F',
It was encoded as the symbol S3 of the local information source, that is, as the 3-bit code word "11r".Moreover, this operation of adding symbol 1 means that the end of the block of the binary symbol sequence is , even when there are no symbols left that cannot be encoded).
In the case of the quaternary local code of FIG. 3, the number of code bits increases by 3 bits due to this operation, but when a higher order local code is used, the number of code bits increases even more. When encoding a binary symbol sequence by dividing it into many blocks, it is necessary to perform the above operation for each block, and the increase in the number of coded bits due to the above operation becomes extremely large, significantly reducing the compression rate. The runner-up is to make it happen. The present invention has been made in order to eliminate the drawbacks of the conventional ones as described above, and provides an encoding device that divides and encodes a symbol sequence from a binary information source into predetermined blocks. If a frequently occurring symbol that cannot be encoded remains at the end of a block, a 1-bit code is assigned to this uncoded symbol sequence for a symbol pattern in which a predetermined number of frequently occurring symbols are consecutive. By assigning the word
It is an object of the present invention to provide an encoding device that can significantly reduce the number of encoded bits in processing the last part of the symbol sequence of each block and achieve high encoding compression.

Hereinafter, one embodiment of the present invention will be described using the block diagram of the facsimile transmitting side apparatus shown in FIG. In FIG. 4, 101 is a memory for storing facsimile signals, 102 is a predictive converter, 103 is a code selector, and 10
4 is an encoder, and 105 is an inserter for identification bits indicating a synchronization code and a local code order.

Here, as compression processing for facsimile signals, consider a method in which the pixel of interest is predicted based on the surrounding pixel pattern, the predicted value and the actual value are exclusive-ORed (called predictive conversion), and then encoded. ing.

The symbol sequence of the binary information source in the above description corresponds to this predictive conversion signal. In general, in facsimile compression processing, a synchronization code is inserted in the middle of a transmission code sequence as a way to deal with errors in code transmission over the line. I'm holding it down.

This synchronization code is made to correspond to facsimile scanning.
Insert every main scan or every few main scans. In this embodiment, a synchronization code is inserted for each main scan, and the above-mentioned operation at the end of the encoded block is performed for each main scan. In addition, the encoding format for encoding the predictive conversion signal is as follows:
Since the predicted coincidence rate varies greatly even within the same stroke, a method is used to select the most efficient one among the 2'' order local codes shown in FIG. 1 for each main scan using the method described below.

The following properties of the 2m-order local code group shown in FIG. 1 are known.

5 When the appearance probability of symbol O is k1 and the appearance probability of symbol 1 is 1-k (k>112), when encoding a binary information source in the format shown in Fig. 1, each of the symbols when the appearance form of the symbol is arbitrary. m, which minimizes the maximum code length for the order, satisfies the following equation.Therefore, by determining m from equation (1), an almost optimal code format can be selected.

Next, the operation of the facsimile transmitting apparatus shown in FIG. 4 will be explained.

When one main scan is completed and facsimile signals 1 for one main scan are stored in the memory 101, reading of the contents of the memory 101 is started.

The output signal 2 of the memory 101 is subjected to exclusive OR of the predicted value and the true value by the predictive converter 102. This output predictive conversion signal 3 is sent to the code selector 103, and the symbol O
(prediction match) and symbol 1 (prediction mismatch) are counted. When the counting for one main scan is completed, the code selector 103
In Equation (1), the code order is determined, and a signal 4 indicating the information is sent to the encoder 104 and the inserter 105.

The inserter 105 receives this and outputs the determined synchronization code and an identification bit indicating the code order. Thereafter, the facsimile signal is read out from the memory 101 again, and the output signal 2 is converted into the predictive conversion signal 3 by the mobile phone converter 102 as before.

Encoder 104 receives this predictive conversion signal 3 and performs encoding as shown in FIG. 1 in accordance with the code format indicated by signal 4 from code selector 103. This code output 6 is sequentially outputted through an inserter 105 (the output signal is 5 on the same output line as the synchronization code and identification bit, and from the memory 101 to 1
If reading of facsimile signals for main scanning is completed and the normal encoding shown in FIG. corresponding) is output. Then, one main scan is performed again, and the same operation is performed thereafter. The receiving device corresponding to the transmitting device shown in FIG. 4 first detects a synchronization code from the received signal sequence, reads the identification bit of the code format that follows it, and determines the code format to be used for decoding.

Subsequent codes are then decoded according to the code format shown in FIG. Then, in the reproduced predictive conversion signal obtained by decoding,
If there is a remainder symbol O that exceeds the number of pixels for one main scan, it can be picked up as invalid. Then, a predicted signal is generated using the already obtained reproduced image signal in the same manner as that performed on the transmitting side, and the reproduced predicted converted signal and this predicted signal are exclusively ORed to obtain the reproduced image signal. Here, if the reference pixel used when performing predictive conversion on both the transmitting side and the receiving side deviates from the effective transmission pixel (such as during predictive conversion of the first pixel in one main scan), this reference pixel is regarded as white. If we use the conventional encoding method for this device,
If the number of scanning lines in one screen is 2400 and the average code degree is 1, then the final processing of each main scan will result in 5 x 2400 = 12
However, according to the present invention, even if 7 unencoded symbols O remain at the end of all main scanning, the number of encoded bits by the same operation is only 2400 bits.

As described above, according to the present invention, in a binary information source encoding device, when a symbol sequence from an information source is divided into predetermined blocks and encoded, an encoding is performed at the end of each block. If there are symbols that occur frequently that cannot be used,
To this uncoded symbol sequence, we assigned a 1-bit code word that is assigned to a symbol pattern when a predetermined number of frequently occurring symbols occur consecutively, so that when processing the last part of the symbol sequence, This has the effect of significantly reducing the number of code bits and achieving high coding compression.

[Brief explanation of the drawing]

Figure 1 is an explanatory diagram showing the code format of the 2m-order local code;
FIG. 3 is an explanatory diagram showing a code format of a quaternary local code, FIG. 3 is an explanatory diagram showing an example of encoding using the code shown in FIG. 2, and FIG. 4 is a block diagram showing an embodiment according to the present invention. be. In the figure, 101 is a memory, 102 is a predictive converter, 1
03 is a code selector, 104 is an encoder, and 105 is an inserter for identification bits indicating the orders of the synchronization code and local code.

Claims (1)

[Claims] 1. In an encoding device that performs predictive conversion on a symbol sequence generated from an information source having two symbols and then divides it into symbol patterns of non-constant length and encodes the symbol sequence, the above symbol sequence of a predetermined length is input and both symbols are encoded. Find the appearance probability of a symbol with a high probability of appearance, k, and find the code order m (m
is a positive integer), and a code selector that inputs the symbol sequence of the predetermined length and selects a code word for a symbol pattern in which 2^m symbols with a high probability of appearance are consecutive according to the code order m. A binary code word with a length of 1 bit is assigned, and one symbol with a low probability of occurrence occurs after i symbols with a high probability of occurrence (i is an integer satisfying 0≦i≦2^m-1) 2^ Binary code words of code word length m+1 bits are assigned to each of the m symbol patterns and encoded sequentially, and the symbol sequence of the predetermined length is divided into each of the symbol patterns in the order of input, and finally 2^m pieces are divided into the symbol patterns. If a symbol with a high occurrence probability of less than and an inserter for inserting an identification bit indicating the code order m determined by the code selector.