JPS6348229B2 - - Google Patents

Description

[Detailed description of the invention]

The present invention relates to a multivalued image signal decoding device that is paired with a multivalued image signal encoding device that compresses data of a multivalued image signal by encoding and reduces transmission time or storage memory capacity. One of the conventional encoding methods for multivalued image signals is predictive encoding. This predicts the currently input image signal x using the input image signal S, and compresses and encodes the prediction error signal e. Various prediction error signals can be considered, but for example, e=x−x^
There is something that says. Here x^ is the predicted value of x. At this time, the higher the probability that the prediction error signal e becomes smaller, the easier it becomes to perform compression encoding, and effective data compression becomes possible. In order to reduce e, it is necessary to determine the optimal x^ for each S. For example, S is the four pixels a, b indicated by circles in FIG.
Let's say it consists of c and d. In FIG. 2, the pixel indicated by a double circle is the image signal of interest x. Second
In the figure, the dotted line indicates the main scanning line of the image, along which the image is scanned from left to right, and when the right end is reached, scanning continues from the left end of the main scanning line below. Assume that each pixel can take N levels {0, 1, 2, . . . , N-1} from 0 to N-1. In other words, let it be an N-value image signal. When the value is N, each pixel a, b, c, d of S is [log ₂ (N-1)] +
It can be expressed with 1 bit. Here the symbol [ ] is [ ]
Let it represent the largest integer not exceeding the number in . S
In general, it is necessary to perform complex nonlinear calculations to determine the optimal x^ that reduces e based on = (a, b, c, d). To achieve this with a circuit,
A ROM (read only memory) that inputs a, b, c, and d as address data and outputs the calculation result x^ as data may be used. This ROM is hereinafter referred to as predictive ROM. When the value is N, a, b, c, d, and x^ are represented by [log ₂ (N-1)] + 1 bits, so the input address of 4 {[log ₂ (N-1)] + 1} bits and [log ₂ (N-1)]+1 bits of output data are required. For example, N=
64, the input address is 24 bits and the output data is 6 bits. In order to construct this predictive ROM using a ROM with a 10-bit address and a 3-bit output, it is necessary to use (6/3)·2 ^24-10 =2 ¹⁵ =32768 ROMs. In order to determine the optimal x^, such a large capacity ROM is required, which is not practical. In order to construct a practical prediction circuit, it is necessary to significantly reduce the required RM capacity. The purpose of the present invention is to obtain near-optimal predictions in small volumes.
It is an object of the present invention to provide a decoding device which is realized by ROM and is paired with a multi-level image signal encoding device which is small in size and efficient. N-value image signal x is divided into n pieces each with at most N ₁ value,
When decoding a coded image signal obtained by decomposing into N _binary , ..., N _o value signals, predictive coding, and compression coding, n types from the first stage to the nth stage from the compression code are used. means for decompressing and decoding the prediction error signals e ₁ , e ₂ , ..., e _o , and converting the already decoded N-value image signal S into a signal _{S 1} of at most L ₁ values, a first-stage predictive decoding means that generates a predictive decoded signal _{x 1 of} at most N ₁ values that roughly specifies x based on the eye prediction error signal e 1; - Based on one x ₁ , x ₂ , ..., x _j-1 at most
A j-th stage that generates a predictive decoded signal x _j of at most N _j values by converting it into a signal S j of L _j values and further specifying x based on S _j and the prediction error signal _{e j of} the j-th stage. Finally, from the n predictive decoded signals x ₁ , x ₂ , ..., x _o obtained by the predictive decoding means from the first stage to the nth stage, N-value image signal x
There is obtained a multivalued image signal decoding device characterized in that it has a means for synthesizing. According to the present invention, multilevel image signals can be stored in a small-capacity ROM.
Thus, it is possible to decode the encoded multi-level image signal obtained by performing almost optimal prediction and encoding. The present invention will be explained in detail below using the drawings. FIG. 1 is a block diagram of an encoding device that is paired with a decoding device of the present invention. The predictive encoding circuit 2 inputs the multi-value image signal 1 obtained by photoelectrically converting the multi-value image, predicts the currently input image signal x based on the reference image signal S, and converts the prediction error signal 3 into a compression code. The signal is sent to conversion circuit 4. An example of the reference image signal S is shown in FIG. In FIG. 2, the dotted line indicates the main scanning line of the image, along which the image is scanned from left to right, and when the right end is reached, scanning continues from the left end of the main scanning line below. a, b indicated by circles,
c and d are four pixels forming the reference image signal S. The x indicated by a double circle is the image signal of interest,
x is the reference image signal S=(a,
b, c, d), and the prediction error is compressed and encoded. In Fig. 2, the number of pixels constituting S is 4, but S may be made up of as many pixels as 5, 66, or a, b, excluding d,
S may be made up of as few pixels as c3. However, if the number of pixels constituting S is too small, the encoding efficiency will decrease, and if it is increased, the scale of the predictive coding circuit will become unnecessarily large. The following explanation will be given assuming that S consists of four pixels a, b, c, and d adjacent to x, as shown in FIG. Returning to FIG. 1, the explanation will be continued. The compression encoding circuit 4 compresses and encodes the prediction error signal 3 and outputs it as a prediction error encoded signal 5. The control circuit 6 sends out clock signals, control signals, and synchronization signals to control each circuit. FIG. 16 shows an example of the predictive encoding circuit 2 for a 64-value multivalued image signal. As already mentioned, S is the 4 pixels a, b,
Suppose it consists of c and d. Of course, by expanding this example, it is easy to construct a circuit for S consisting of a large number of pixels and a multivalued image signal having even more levels. As already explained, 64
In order to obtain the optimal prediction signal x^ for x when the reference image signal consists of 4 pixels, a ROM with 24 bits of input address and 6 bits of output data is required. It is not very practical to use such a large capacity ROM. To avoid this, the predictive encoding circuit shown in FIG. 16 performs predictive encoding of a 64-value image signal by decomposing it into two stages of predictive encoding of an 8-value signal. Of course, as will be explained in detail later, the 64 decomposition method includes 16 values in addition to 8 values x 8 values.
Various values can be considered, such as value x 4 values, 4 values x 4 values x 4 values, 8 values x 4 values x 2 values, etc. Convert 64 signal values to 0, 1,
2, ..., 63, the first stage predictive coding is such that the image signal of interest x is x<8, 8x<16, 16x<
24, 24x<32, 32x<40, 40x<48, 48
It performs predictive coding to determine whether x<56 or 56x, and the second stage predictive coding is performed after the first stage predictive coding, and finally calculates the signal value of x. is specified. For example, if the first stage predictive encoding is x<8, then 0, 1, 2, 3,
Predictively encode which of the 8 ways x is 4, 5, 6, and 7, and if 8x<16, 8, 9,
Predictive coding is performed to determine which of the eight options 10, 11, 12, 13, 14, and 15 it is. It is most efficient to use S = (a, b, c, d) to perform the first stage predictive encoding, but the input address is 6 bits x 4 = 24
It is not practical because a large capacity ROM of bits must be prepared. Therefore, in this predictive encoding circuit, S=(a,
b, c, d) shown in Table 1 T _1,1 = 8, T _2,1 =
16, T _3,1 = 24, T _4,1 = 32, T _5,1 = 40, T _6,1 = 48,
The 8-value signal S ₁ = (a ₁ , b ₁ , c ₁ , d ₁ ) is converted into an 8-value signal using 7 threshold values of T _7,1 = 56, and this S ₁ is used to predict the first stage. Do the following. In this way, a ROM with an input address of 3 bits x 4 = 12 bits is sufficient. If x is 8-valued using the above seven thresholds T _1,1 , T _2,1 , ...T _7,1 and x ₁ is the result, the first stage prediction is S ₁ and x ₁
It is a prediction. The relationships between x and x ₁ and a and a ₁ are shown in the first and second columns of Tables 3 and 4. b and
The relationships between b ₁ , c and c ₁ , and d and d ₁ are the same as the relationships between a and a ₁ in Table 4. As with the first stage, the second stage of prediction is most efficient if S = (a, b, c, d) is used, but a large capacity ROM with an input address of 6 bits x 4 = 24 bits is prepared. It is not practical because it has to be done. Therefore, in this predictive encoding circuit, S=(a, b,
c, d) as T _1,2 <T _2,2 <T _3,2 <T _4,2 <T _5,2 <T _6,2 <
Select 7 thresholds of T _7,2 and use these to determine S=(a,
b, c, d) and converts them into 8-value signals S ₂ = (a ₂ , b ₂ ,
c ₂ , d ₂ ) and performs second-stage predictive encoding using this S ₂ . In this way, a ROM with an input address of 3 bits x 4 = 12 bits is sufficient. 7 thresholds T _1,2 ,
T _2,2 , ..., T _7,2 are determined based on the value of x ₁ as shown in Table 2, and the second stage predictive coding is efficient by S ₂ .

【table】

【table】

【table】

Claims (1)

[Claims] 1 Decompose the N-value image signal x into n (>2) signals of at most N ₁ value, N ₂ values, ..., N _o values, perform predictive encoding, compression encode, and decode the resulting encoded image signal. When decoding, a means for decompressing and decoding the prediction error signals e ₁ , e ₂ , ..., e _o from the first stage to the nth stage from the compression code, and a means for decoding the already decoded N-value image signal S are used. A first step that converts into a signal S ₁ of at most L ₁ values, and generates a predictive decoded signal x ₁ of at most N ₁ values, which roughly specifies x based on S ₁ and the first stage prediction error signal e ₁ .
The predictive decoding means in the second stage converts S into a signal S j of at most L _j values based on j-1 pieces of x ₁ , x ₂ , ..., X _{j-1 for} j from 2 to n _. At most, x can be specified more precisely based on the j-th prediction error signal e _j .
n predictions obtained by the _j -th stage predictive decoding means that generates the predictive decoded signal x _j of N j values, and the predictive decoding means from the first stage to the n-th stage. Using the decoded signals x ₁ , x ₂ , ..., x _o , convert x ₁ , which roughly specifies x, into a value that specifies x in detail using x ₂ , and further convert the detailed specified value to x ₃ . 1. A multivalued image signal decoding device characterized by having means for reproducing x in detail by repeating this operation up to x _o and finally reproducing x.