JPS637050B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a method for encoding image signals and the like.

A number line representation encoding method is known in which the output sequence of an n-valued image signal is mapped onto a number line from 0.0 to 1.0, and the coordinates expressed in n-ary form are used as code words to encode the output sequence of symbols. ing.

For the sake of simplicity, the case where n is 2 will be described below.

Figure 1 shows its outline. Now, let the appearance probability of symbol 1 be r, and the appearance probability of symbol 0 be (1-r). When the output sequence length is 3, the coordinates of each of the rightmost C (000) to C (111) are expressed in binary and truncated to a distinguishable digit to form each code word. It is possible to decode it by going through the same process as on the other side.

In such an encoding system, the initial values of Ai and Ci are set to 1.0 and 0.0, and the coordinate Ci for the output sequence is determined by the following calculation.

When output symbol a _i is 0 Ai = (1-r) A _i-1 Ci = C _i-1 When output symbol a _i is 1 Ai = rA _i-1 Ci = C _i-1 + (1-r )A _i-1 Now, when creating such a device, it is of course necessary to determine the calculation precision in advance, and when calculating Ai, the first digit below the decimal point is The numbers below the specified number of digits will be rounded down.

Regarding how to determine this calculation precision, if the number of significant digits is set small, Ai will be estimated to be smaller than the actual value,
As a result, the number of digits to represent Ci increases, and conversely, if the number of significant digits is set to a large number, the number of digits to represent Ai increases, which increases the number of digits to represent Ci.

The present invention was made to eliminate the drawbacks of the conventional ones as described above, and it is possible to encode the sequence Si without necessarily using the coordinates Ci from Ci to Ci +
Focusing on the fact that correct decoding is possible using any coordinate position up to Ai, after representing Ci + 1/2 Ai in n-ary, the code can be decoded by using up to a predetermined number of digits calculated from Ai as a code word. The maximum length is
The purpose of the present invention is to provide an encoding method that guarantees that the amount of information is kept below a certain value determined by the amount of information, and that the higher the calculation precision, the higher the efficiency.

Now, given the symbol sequence Si, sequentially
We will calculate Ci and Ai, and finally we will calculate Ci and Ai.
How many digits below the decimal point should be used as a code word after calculating +1/2Ai will be explained below. First, find an integer value m _i that satisfies the following formula.

1/2 ^mi+1 Ai<1/2 ^mi , that is, if Ai is expanded in binary, it becomes as follows.

Next, consider S′i as a series different from Si. Similarly, finding m′i, we get The codeword of Si is mi + 2 bits, the codeword of S′i is m′i
Take +2 bits. At this time, let Ci > C′i.

If mim′i, in order to distinguish between the two codewords, (Ci+1/2Ai)−(C′i+1/2Ai)1/2 ^m ′ ⁱ⁺² (
1) The following conditions are necessary.

Also, in the case of m′imi, in order to distinguish between the two code words, (Ci+1/2Ai)−(C′i+1/2A′i)1/2 ^mi+2 (
2) must be satisfied.

However, from CiC′i+A′i, (Ci+1/2Ai)−(C′i+1/2A′i)1/2(A
′i+A′i) 1/2 (1/2 ^mi+1 +1/2 ^m ′ ⁱ⁺¹ ) Therefore, equations (1) and (2) are satisfied. Therefore, the series Si is
Even if the sequence S'i is encoded using m'i+2 bits and mi+2 bits, the two can be distinguished.

It is also shown that the relationship between the average code length L and the entropy H is as follows.

L=ΣAi(mi+2)H+2 Therefore, it is sufficient to expect the average code length to be about 2 bits larger than the entropy, and according to this encoding method, it is possible to encode image information etc. with high efficiency.

Next, code words for a specific example are shown in FIG. However, r=1/4. Furthermore, in a system that preferentially decodes the synchronization signal, the length of the code word for image data is also known, so it is possible to omit the continuous "0" signal at the end of the code word.

Note that the present invention is also effective for encoding electrical signals other than image information.

As described above, according to the present invention, Ci + 1/2 Ai is expressed in n-ary and then the number of digits calculated by a predetermined formula is taken as a code word using this n-ary expressed Ai. The required number of digits of the codeword can be obtained independently of precision, and high efficiency is guaranteed.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing the principle of number line expression encoding, and FIG. 2 is a diagram for explaining the encoding method according to an embodiment of the present invention. A(0), A(1) to A(111), C(0), C(1
)
~C (111)...Variable.

Claims (1)

[Claims] 1 Symbols of n-value information source {0, 1, 2, ..., n
-1} is mapped to a point between 0.0 and 1.0 on the numerical line, and the n-ary coordinates of this point are used as the code word. In this encoding method, the following two variables are used:
Ai, Ci Ai=r _j A _i-1 (However, when j = 0, Ci = Ci _-1 ) (Here, r _j is the probability of appearance of symbol j, and the initial values of Ai and Ci are 1.0 and 0.0, respectively.) are calculated one after another with a predetermined precision, Finally, find the integer value mi that satisfies the condition 1/n ^mi+1 Ai<1/n ^mi , truncate the n-ary coordinates of Ci+1/2Ai to the decimal point (mi+2) bits, and select the series of the (mi+2) bits. An encoding method characterized in that the output sequence is transmitted as a code word of the output sequence. 2. The encoding method according to claim 1, characterized in that when there is a continuous "0" signal at the end of a code word, the signal is omitted and transmitted.