JPS61205093A - Device for high efficiency coding of color image - Google Patents

Abstract

PURPOSE:To match excellently inter-channel difference vector quantization, and to attain high speed transmission of a high quality color image superior in color reproducibility at an extremely low bit rate by utilizing the correlation between color signals to generate a predicting signal. CONSTITUTION:Each color image difference signal is outputted from a subtractor subtracting from each channel for the color image correlation predicting signal formed with the use of the encoded color image reproduction signal system from the color image input signal system consisting of three components. These color image difference signals are vector-quantized in the K-order signal space. The output vector giving the lowest distortion to the input vector, a vector quantizer outputting its index and the output vector giving the smallest distortion are added as the color vector quantization signal to the color image correlation predicting signal. There is also provided an encoding part composed of an adder for obtaining the respective color image reproducing signal, and a color image signal predicting unit forming the respective color correlation estimation signal by multiplication of the respective color image signal predicting coefficient and the color image signal interval prediction coefficient relative to the coded signal system of the respective color image reproducing signals.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a high-efficiency encoding device for efficiently transmitting an image by removing redundancy from a color image signal consisting of three components.

[Conventional technology]

FIGS. 4 and 5 are configuration diagrams of an encoding section and a decoding section showing an embodiment of a conventional color image encoding device.

In Figure 4, 141, 151 and (61 are 3
A subtracter that subtracts a color image prediction signal for each component from a color image input signal consisting of components and outputs a color image difference signal, and α3 collectively vector quantizes the three component color image difference signals in a three-dimensional signal space. A vector quantizer that outputs a color image difference vector quantization signal and a vector quantization index.

34.Q8, 119 and (2) are adders that add the color image difference vector quantized signal and the color image prediction signal for each component to obtain a color image prediction signal; 34. (c) and (to) are an R predictor that forms a color image/predicted signal for each component from the color image reproduction signal sequence;
A predictor and a B predictor.

In FIG. 5, (42) is a vector quantization decoder which receives a vector quantization index and outputs a corresponding color image difference vector quantization signal. In the drawings, parts with the same reference numerals indicate the same or equivalent parts.

Next, the operation of the encoding section will be explained according to FIG.

Regarding a color image input signal series consisting of three color components of red, green and blue, the red image input signal (11f R(z)).

Let the recorded image input signal [21' (z G (t) and the blue image input signal (3) be B). t-0, 1, 2,
...... indicates the signal sequence number.

Now, each red, green and blue image prediction signal (45), (
44) Oj U (45) t”k tL Tsure PR (
t), PG (j), Oyohi PB C Ritoshi,
Each color image difference signal (46), (47) and (
48) are respectively imported (t), CG (B) and CB (
t), and further convert each color image difference vector quantized signal (50), (51) and (52) into face CI), 'ja (4 and ?BΩ), respectively.

In addition, each color image reproduction signal (53) (54) and (
55)'tR(t), G(t) and BCt).

At this time, the encoding section shown in FIG. 4 executes the following arithmetic processing.

εR(4-R(et)-pR(z), εR(t)-ε
R(z) + QR(z)εo(t)-GΩ) -PG
(t), (t) = εa (t) + C0 body) g
BGc)-B(z)-pB(z), gB(z)-g
B (t) + qB(z)R"-PR(z)+?R(4
= R (z) + QR (4G (t) -PG (4+
Ministry () (j) −G (t) + QG (t)'m
a>-pBa>+qBrt>= B<t)+Q,<t
>PR(t)! 1lICR fist R(t-1)PGΩ) −
CG-G(t-t) PB(t)wCBIIG(t one here e C1(e CG and CB are prediction coefficients for each channel of each color image signal in predictive coding.tta, QR(4, Qe (Wt) Ojhi QB <
j) Vector quantization noise for each color image signal channel generated by the tl/tor quantizer a3.

The vector quantizer (13 processes the color image difference signal in the following steps. Now, the three-dimensional signal space S shown in FIG.
5, the color image difference signal εRK).

εa (t) and εBl,j)'k are collectively referred to as input vector D. This input vector quantizer produces N output vector quantities j (j-1, 2,......N
), and select the output vector that minimizes the distance d(x, yj) between the input vector 5 and the output vector to obtain the color image vector quantized signal εR(/, ),
εa (4 and CB(t)), and the index 1 of the output vector amount 1 that produces the minimum distortion is sent to the decoding unit as encoded data (49). That is, if vector quantization processing is VQ, the (xl, x2.x5)t=(εR(t),CG(t
), CB(4) )'qj−(Yj + * Yj 2
*Yj 5 )d(and yj)-you(XK-YjK
) K-+ VQ: ll→yi if d(31, yi)
A color image difference vector quantized signal is obtained as <dQl, yj) for auj ヱ1 (yit*yi2syia)1 savingRCt)βG), ↑8(t)), and encoded data 1 is sent to the decoding unit. Output.

Next, the operation of the decoding section shown in FIG. 5 will be explained.

First, the vector quantization decoder (42) calculates the minimum distortion mi for the encoded data (49), that is, the input vector xt.
It receives the intex 1t- of the output vector quantity 1 which becomes x d(x4yj
Similar to the vector quantizer αj in the encoder shown in the figure, the vector doji conversion decoder (42) in the decoder shown in FIG.・・・
・) This can be easily achieved by making all the preparations. By mapping transformation from this index to the output vector L1, each color image vector quantized signal εR1εG and ? 8 is output to the signal lines (50), (51) and (52). A reproduced color image signal sequence is obtained by adding this signal to a color image prediction signal formed from each successively reproduced color image signal.

[Problem that the invention seeks to solve]

Since the conventional color image encoding device is configured as described above, even though red, green, and blue image signals are subjected to differential vector quantization, a high data compression rate can be obtained.
There is a problem in that the accumulation of vector quantization noise between each color image signal in the encoding loop affects each other and is not averaged, resulting in a severe color shift that appears at the 0 image change point.

This invention was made in order to solve all of the above-mentioned problems, and it is possible to smooth the difference vector quantization noise between all color image signal channels in the encoding loop, and also to smooth out the sharp changes in the six colors. An object of the present invention is to obtain a highly efficient color image encoding device capable of improving color reproducibility in a region without reducing encoding efficiency.

[Next method to solve the problem]

A color image high-efficiency encoding device according to the present invention.

Vector quantization noise is eliminated across each channel by multiplying each reproduced color image signal of the encoding unit by a prediction coefficient that utilizes the correlation between channels to form a color image prediction signal for each channel. Smoothing is performed to prevent the accumulation of noise in the encoding loop.

[Effect]

Therefore, the method of generating a color image prediction signal in this invention is well matched with a vector quantizer, prevents expansion of vector quantization noise due to noise interference between color image signals, and improves the color reproducibility of reproduced images. This has the effect of completely preventing color leakage in changing areas and allowing high-quality images to be encoded with low bits.

[Embodiments of the invention]

A color image high-efficiency encoding device which is an embodiment of the present invention will be described below with reference to the configuration diagrams of the encoding section in FIG. 1 and the decoding section in FIG. 2.

Figure 1 1c Oite, +41, +51, +61
, (13, Ql, fil, W.

(Goods), (C) and @ are the same as or equivalent to those in Figure 4. (to) to @ is a multiplier that multiplies the inter-channel prediction coefficient of each color image signal; This is an accumulator for each channel of the signal multiplied by the prediction coefficient.

In FIG. 1, (42) is a vector quantization decoder;
Other parts are the same as those in FIG. 1 with the same reference numerals.

The operation of the color image high-efficiency encoding device according to the present invention will be explained below, starting from the encoding section in FIG.

Now, regarding each color image input signal consisting of red, green, and blue, the red image input signal (11tR(4, recorded image input signal 1
2+ total G (/,), blue image input signal 13+'iB
ρ).

t-t, 2. . . . indicates the time series number of each signal.

At this time, each color image correlation prediction signal +71 +81 and (9) of the red, green, and blue channels are calculated as p(4).

p3 rt> and p; g) and each color image difference signal obtained by subtracting each color image correlation prediction signal from each color image input signal through a subtracter +41151 and (6) for each channel. [11, The three signals (lυ and (12'i, respectively gH body), tta (t) and εB (4, and the above-mentioned εR(1) *εG (j) and εB body) in the three-dimensional signal space 83 Each color image vector quantized signal α9a[9 and αη, respectively?] vector-quantized by the vector quantizer αj.
, (z), 9G (z) and jB<t). Further, each color image vector quantized signal and each color image correlation prediction signal are inputted to an adder I for each channel.
To each color image obtained by adding with J & a9■
The raw signals cnw and (to) are converted to R(j) and ()(
4 and B (/-), and based on each of the color image reproduction signal sequences, the R predictor 04, the Ck predictor (c), and the B predictor 04 utilize the color image signal sequence for each channel. Each color image prediction signal @ (to) and @ all PR (
t)*Pa (4 and FB(/,).

At this time, the encoding section shown in FIG. 1 executes the following arithmetic processing.

Axis (j)-RK)-P%(4,'H(t)-tR(t)
+ q; (z)εau>-ta(t>-FS Gri
, ? o(j)=gG(4)+4GriεB(,t
) = εB (4 - time (t) 5) - εB (t) + Q;
3 (t) Zen(4-pkan(t)+ ``ritsu)-R(4+
Q; (t) @(t-1-p3 (t) + ?0 (
1ri-G (j) + q5 (z)9(z)-p;(
t)+fB(z)-B(z)+q2(z)PC)-
CR-R (t-rePC body)-c. -G(t-t) p (t) = cB-B(z-Q B P; (') = a 11°PR't) + alz
oPG k) +a+56PB(t)p, (1) −
a2+ ・PR(t)+ az2・PG (j) +
a2! l・PB (1rip2(z) = a51-PR
(4+ as2・PG Gri + ass・PB (
t) Here at+ @ a12 m als * &2
1 + a22 * a25 * a31 s a shun *
als are inter-channel prediction coefficients derived from the correlation between each color image signal. Regarding the above coefficient, a11 + a12 + a15 < 1 a21 +
a22 + a25 <1a5+ + &s2+ a
It is desirable to preserve the relationship ss <, 1.

The above-described means for forming a color image correlation prediction signal for each channel by multiplying it by a coefficient corresponding to the correlation between adjacent pixel color signals according to the present invention smoothes the vector quantization noise generated by the vector quantizer α3 and encodes the color image correlation prediction signal for each channel. Cumulative propagation in the conversion loop can be prevented. That is, Σ[q;(z)'+q3(t)2+q;(z)']<Σ
(q signal is 10QG Ct)2+oB (ri).

The vector quantization procedure is the same as that shown in the explanation of FIG.

The operation of the decoding unit shown in FIG. After conversion to 6, each color image reproduction signal 'R<t), ';B (1) and 9 (
2) Get +. As shown in the encoder section, vector quantization noise is reduced and improved compared to conventional devices.

Note that in the above embodiment, a predictor for intra-frame predictive coding,
Although we have shown the case where adaptive control of the vector quantizer is not included, what can also be implemented in the case of intra-frame or inter-frame predictive coding, or even adaptive switching control of the predictor and vector quantizer. Of course.

Further, in the above embodiment, high-efficiency encoding of red, green, and blue color image signals has been described, but the present invention can be similarly applied to a color image signal that is a combination of a luminance signal and two color difference signals.

〔Effect of the invention〕

As described above, the present invention utilizes all the correlations between color signals to form a predicted signal, which matches well with the inter-channel difference vector quantization, and achieves high resolution with excellent color reproducibility at an extremely low bit rate. This has the effect of allowing high-speed transmission of quality color images.

[Brief explanation of the drawing]

FIG. 1 is a configuration diagram of an encoding section of a color image high-efficiency encoding device according to an embodiment of the present invention, and FIG. 2 is a configuration diagram of a decoding section of a color image high-efficiency encoding device according to an embodiment of the present invention. 3 is a conceptual diagram of vector quantization of a color image difference signal according to the present invention, FIG. 4 is a block diagram showing an embodiment of the encoding section of a conventional color image encoding device, and FIG. FIG. 2 is a configuration diagram showing an example of a decoding section of a conventional color image encoding device. In the figure, +41, 151 and (6) are subtractors, (1
3 is a vector quantizer, 0υ, (19 and ■ are adders, 24 is an R predictor, (to) is a G predictor, 弼 is a B predictor, C3G to (to) is a coefficient multiplier, C(l , f4
G and (4-key accumulator, (42) are vector quantization decoders. In Figure 9, parts with the same reference numerals indicate the same or equivalent parts.

Claims (1)

[Claims] a subtracter for each channel to subtract a color image correlation prediction signal formed using the encoded color image reproduction signal sequence from a color image input signal sequence consisting of three components and outputting a color image difference signal for each color; A vector quantizer that vector quantizes image difference signals into K units (K is an integer of 3 or more) in a K-dimensional signal space and outputs an output vector and its index that cause minimum distortion with respect to the input vector, the minimum an adder that adds the output vector that becomes distortion as a color image vector quantized signal to the color image correlation prediction signal for each channel to obtain each color image reproduction signal; In contrast, the color image signal predictor is configured to form each color image correlation prediction signal by multiplying each color image signal prediction coefficient and each color image signal inter-prediction coefficient based on the correlation not only between adjacent pixels but also between each color channel pixel. 1. A color image high-efficiency encoding device, characterized in that it is equipped with an encoding section.