JPS63177672A - Picture encoding transmission system - Google Patents

Abstract

PURPOSE:To suppress the degradation of picture quality and reduce the quantity of generated information by roughening the quantization characteristic of variance to reduce the number of quantizing levels if the absolute value of the average value of a difference vector signal is large. CONSTITUTION:A input picture signal is divided into blocks and is converted to a vector signal, and a reference block is generated by a motion compensating part 2, and the difference between the vector signal and a selected vector signal is calculated by a subtractor 5 to output a difference vector signal to a vector quantizing and encoding part 3. An average value (m) and a variance sigma of the difference vector signal are calculated by a operating part 3a, and it is decided whether they are valid or not by an validity invalidity deciding circuit 3b. The average value (m) and the variance sigma are quantized, and the quantization characteristic for variance is roughned if the absolute value of the average value (m) is large, thereby reducing the quantity of encoded signal as the result of variable length encoding.

Description

DETAILED DESCRIPTION OF THE INVENTION [Industrial Application Field] The present invention relates to an image encoding and transmitting method, particularly to an improvement of an image encoding and transmitting method for video conferences, video telephones, and the like.

[Prior Art] In general, image transmission used for video conferences, video telephone calls, etc. involves a huge amount of image information. A compression encoding method that reduces (compresses) the amount of data has been put into practical use.

Among these methods, vector quantization methods including motion compensation are known as methods for compressing the amount of information with a high degree of compression.

Hereinafter, a conventional image encoding and transmission method will be explained based on the drawings.

FIG. 1 shows the basic configuration of an image coding and transmitting apparatus to which an interframe coding method using a vector quantization method including motion compensation is applied.

As shown in the figure, the image encoding and transmitting device is roughly divided into a preprocessing section (1), a motion compensation section (2), and a quantization/encoding section (3). There is.

The preprocessing unit (1) is a circuit that divides an image input signal (100) into blocks of adjacent pixels on the image and creates a dimensional vector signal (101) for each block, and a frame memory (4). A blocked image signal of one frame before the current image signal is stored.

The motion compensation unit (2) refers to a current vector signal (101) and a plurality of blocks including a block stored in the frame memory at the same position on the image as a block corresponding to the current vector signal (101). A reference block generation unit (2a) that creates a block and calculates block position information (102al) and a vector signal (102a2) of the reference block, and distortion between the current vector signal (101) and the reference vector signal (102a2). (for example, Euclidean distortion or absolute value distortion), and a distortion calculation unit (2b) that performs calculations and selects a block with the minimum distortion from among the reference blocks.

A subtracter (5) performs a difference operation between the current vector signal (101) and the vector signal (102b2) of the selected block selected by the motion compensation unit (2), and converts the difference vector signal (105) into a vector quantum and sends it to the converter/encoder (3).

The vector quantization/encoding unit (3) includes a calculation unit (3a) that calculates the average value (m) and variance (σ) of the difference vector signal (105), and a calculation unit (3a) that calculates the average value (m) and variance (σ) of the difference vector signal (105). σ
) and a threshold value (Thl) for controlling the amount of information compression;
(Th2), a valid/invalid determination circuit (3b) that determines whether the selected block is valid/invalid, and a differential vector signal (
105), a codebook (3d) in which patterns of a plurality of normalized image vector signals are stored; a vector quantization unit (3e) that selects a pattern that is the same as or similar to the normalized difference vector signal (103c) normalized in step 3C);
m) and a quantization unit (3f) that encodes the variance (σ);
It consists of a variable length encoder (3g) that encodes the selected pattern number and the quantized average value and variance.

Next, the signal flow will be explained.

First, an image input signal (100) is divided into blocks by a preprocessing section (1) and converted into a vector signal (101).

Then, the motion compensation unit (2) generates a reference block, selects the block with the least distortion from the vector signal (101) among the reference blocks, and selects the selected block position information (102bl) and the selected vector signal. (102
b2) is output to the subtractor (5).

Then, in the subtracter (5), the vector signal (101
) and the selection vector signal (102b2), and a difference vector signal (105) is output to the quantization/encoding section (3).

Then, in the quantization/encoding unit (3), the difference vector signal (105) is used to calculate the average value (m) and deviation value (σ) of the difference vector signal (105) in the calculation unit (3a). , in the valid/invalid discrimination circuit (3b), an average value threshold (Thl) and a deviation value threshold (Th
2), the validity/invalidity is determined as shown in the following equation.

If the determination is invalid, it is assumed that the current block and the selected block are the same, and only the selected block position information (102bl) and invalid information are encoded and temporarily stored in one transmission buffer (6). .

On the other hand, if it is valid, the difference vector signal (105) is normalized by the normalization unit (3C) as data to be transmitted based on the following equation.

y, -(xl-m)/σ (however, 1-1.2...k) and the normalized normalized difference vector signal (103
In c), the vector quantization unit (3e) selects a pattern most similar to the normalized difference vector signal (103c) from the codebook (3d).

Further, the quantization unit (3f) quantizes the mean value (m) and variance (σ), respectively, as shown in FIG. and,
The pattern number and selected position information (1
02bl), the information that it is valid, the quantized mean value (m) and deviation value (σ) are sent to a variable length encoder (3g
) and temporarily stored in a transmission buffer (6).

The encoded image signal temporarily stored in the transmission buffer (6) is transmitted frame by frame.

On the other hand, the thresholds (Thl) and (Th2) are controlled by the amount of image encoded signal stored in the transmission buffer (6) of the previous frame, and when the amount of signal is large, the value is large, and the amount of encoded signal is When there is little, it is controlled to be a small value,
The degree of compression is controlled for each frame.

[Problem to be Solved by the Invention 1] As described above, according to the conventional image coding and transmission method, for images that change rapidly, there are problems such as data cutting using a threshold value and compression of coded information using variable length coding. However, the amount of encoded information cannot be reduced sufficiently, and as a result, a frame-dropping method is used in which the entire screen is deleted without being transmitted, resulting in poor screen continuity and visual impact. However, there was a problem in that the feeling of discomfort increased.

The present invention has been made to solve the above-mentioned problems, and an object of the present invention is to obtain a coding transmission method that improves the degree of compression of the amount of information in a vector quantization interframe coding method. .

[Means for Solving the Problems] In order to achieve the above object, as described above, the image encoding and transmission method according to the present invention uses a vector quantization method, and a differential block encoding method. Normalize the vector signal, select the pattern closest to the normalized difference vector signal from a codebook in which normalized patterns are stored in advance, and quantize and encode the mean value (m) and variance (σ), respectively, and select the pattern. The pattern number is sent to the transmission buffer along with the specified pattern number. At the time of decoding, we focused on the fact that it is decoded in the following form, △ ε■σy+m Δ (where ε is the decoded difference signal vector y is the normalized pattern vector). Taking advantage of the fact that when the absolute value (1ml) is large, even if the quantization distortion of the dispersion is larger than usual, it is not noticeable in the image, information is generated by coarsening the quantization of the dispersion (σ) and reducing the number of quantization levels. This is to reduce the amount.

[Operation] As explained above, according to the image coding transmission method according to the present invention, the absolute value (
1 ml) is large, it is possible to obtain an image coding and transmission method in which the amount of information generation can be effectively reduced by roughening the quantization characteristic of the variance (σ).

[Embodiment] Hereinafter, a preferred embodiment of the image encoding and transmission method according to the present invention will be described based on the drawings.

The basic block configuration of an image encoding and transmitting apparatus to which the method of the present invention is applied is shown in FIG. 1, as in the prior art.

The features of this embodiment are the quantization/encoding unit (3)
The quantization unit (3f) performs quantization encoding of the average values (m) and (σ) provided in the quantization unit (3f), and as shown in FIG. A) Multiple quantization characteristics are prepared, and the switching of the dispersion quantization characteristics is selected based on the absolute value (1ml) of the average value (m), and when the absolute value (1ml) is large, the dispersion quantization characteristics are The roughening characteristics are chosen to be coarse. Next, the flow of signals will be explained.

First, an image human input signal (100) is divided into blocks in a preprocessing section (1) and converted into a vector signal (101).

Then, a reference block is generated in a motion compensation unit (102), and the vector signal (101) in the reference block is
The block with the smallest distortion is selected, and the selected block position information (102bl) and the selected vector signal (102bl) are selected.
b2) and information that it is invalid are output to the subtractor (5).

The thresholds (Thl) and (Th2) are controlled by the amount of encoded image signals stored in the transmission buffer (6) of the previous frame, and are set to large values when the amount of signals is large, and when the amount of encoded signals is small. The time is controlled to a small value.

Then, in the subtracter (5), the vector signal (101) is
and the selected vector signal (102), and a difference vector signal (105) is output to the vector quantization/encoding section (3).

Then, in the vector quantization/encoding unit (3), the difference vector signal (105) is converted to the mean value (m) and variance (σ
) is calculated, and the valid/invalid discriminating circuit (3b) calculates the average value required threshold value (Thl) and the dispersion threshold value (Thl).
2), the validity/invalidity determination is performed as shown in the following equation.

If the determination is invalid, the current block is considered to be the same as the selected block, and only the selected block position information (102b) and the information that it is invalid are encoded and stored in the transmission buffer (6). Memorize temporarily.

On the other hand, if it is valid, the difference vector signal (105) is normalized as transmission data in the normalization section (3c) based on the following equation.

Y 1- (x t m)/σ (however, 1-1.2...k) and the normalized normalized difference vector signal (103c
), a vector doji conversion unit (3e) selects a pattern most similar to the normalized difference vector signal (103C) from a codebook (3d) in which normalized patterns are stored in advance.

In addition, the average value (m) and variance (σ) are calculated by the quantization unit (3
quantized in f). At this time, the selection of the quantization characteristic for dispersion in the quantization unit (3f) is determined by the absolute value (1 ml) of the average ti (m),
When is large, the quantization characteristics are coarse and the quantization level is selected to be small.

Then, as transmission information, the pattern number, selected position information (102bl), the quantized mean value (m), and variance (σ) are each variable-length coded, and the transmission buffer (6
) is temporarily stored.

The image encoded signal temporarily stored in the transmission buffer (6) is transmitted frame by frame.

According to this embodiment, when the average value (m) and the absolute value (1 ml) are large, the amount of encoded signals resulting from variable length encoding is reduced by roughening the dispersion quantization characteristics.

In addition, the method of the present invention is based on the average value (m) of the differential vector signal.
Since this method coarsens the quantization characteristics based on the absolute value (1 ml) of , the encoded signal amount can be deleted without significantly affecting the image quality.

In the above embodiments, an example of an image to which a vector quantization method including motion compensation is applied is shown, but the same effect can be obtained if an inter-frame difference method using a vector quantization method is used.

[Effects of the Invention] As explained above, according to the image coding transmission method according to the present invention, the absolute value (
By roughening the dispersion quantization characteristics according to the size of 1 ml), it is possible to obtain an image encoding and transmission method that can minimize the deterioration of image quality caused by the coarse quantization characteristics and suppress the generation of information amount. can.

[Brief explanation of the drawing]

FIG. 1 is a basic block diagram of a preferred embodiment of an image encoding and transmitting apparatus to which the image encoding and transmitting method according to the present invention is applied, and FIG. Detailed block diagram of the quantization unit in the apparatus. FIG. 3 is a detailed block diagram of the quantization unit in the image encoding and transmission apparatus to which the conventional image encoding and transmission method is applied. In the figure, (3) is a vector quantization/encoding unit;
a) is an arithmetic unit, (3b) is a valid/invalid discrimination circuit, (3c
) is the normalization section, (3e) is the vector quantization section, (3f)
is the sonification department. The same reference numerals in the figures indicate the same or corresponding parts.

Claims (2)

[Claims] (1) A pre-processing step in which a plurality of pixels at close positions on the image of the image input signal are grouped together into blocks, and a vector signal is created and output for each block; Read the vector signal of the previous frame block located at the same position as the current input block from the frame memory in which vector signals for one image are stored, and calculate the difference vector signal by calculating the difference between the input block and the previous frame block. calculating the average value and variance of the difference vector signal, and determining that the input block and the previous frame block are the same if the average value and variance are within a threshold range; is,
The invalid signal and position information of the input block are encoded, the encoded signal is sent to the transmission buffer, and if the mean value and variance of the difference vector signal are outside the threshold range, the difference vector signal is normalized and the normalized difference is generated. A signal is obtained, a normalization pattern close to the normalized difference signal is selected from a codebook that stores multiple types of normalization patterns in advance, and the effective signal, normalization pattern number, average value, and variance are quantized and encoded. , an encoding processing step for sending to a transmission buffer; the threshold value is controlled by the amount of encoded signal temporarily stored in the transfer buffer, and when the amount of encoded signal is large, the threshold value is increased, and the encoding process is performed. In an image encoding transmission method using a vector quantization method comprising: a threshold control step in which the threshold value is decreased when the signal amount is small; An image encoding and transmitting method characterized in that when the absolute value of the value is large, the amount of encoded signal is reduced by roughening the quantization characteristic of the variance and reducing the number of quantization levels. (2) The difference vector signal calculation step includes a motion compensation processing step of selecting a block for motion compensation, and a step of calculating a difference vector signal by calculating a difference between the input block and the selected block. The image encoding transmission method according to item (1).