JPH05227547A - Method and system for coding and compressing video signal - Google Patents

Abstract

PURPOSE: To compress an information amount transmitted at the time of coding video signals and transmitting them, through an inter-system cable or the like and to reduce a bit rate to a value allowed by the channels by providing a specified encoder-compressor. CONSTITUTION: Video signals are digitized and coded to blocks of two kinds of pixels, a bit map and a frame map. An encoder 14-a compressor 18 code A/D converted video signals to the blocks of a matrix, composed of the plural pixels and the bit map for defining first and second reference values and find, specify and code non-redundant information inside the coded block. Then, the respective blocks are compared with the corresponding block of a frame immediately before, and in the case of a new block, the block is compressed, the bit map is eliminated and a present color value is coded in the form of a difference from the color value immediately before. In such a manner, digital data required for indicating the video signals are reduced.

Description

Detailed Description of the Invention

[0001]

FIELD OF THE INVENTION This invention relates generally to processing information signals, and more particularly to information transmitted from an encode site to a decode site, such as when transmitting a video signal over an intersystem cable or telephone line. It relates to the processing of time series information signals, such as video signals for compressing quantities.

[0002]

The prior art is US Pat. No. 4,816,901 (3 Mar. 1989) by the inventor of the present application and others relating to "Method and System for Compressing Color Video Data".
U.S. Pat. No. 4,847,677 (July 1989) by the inventor of the present application regarding "Video Communication System and Method for Compressing Digital Color Video Data" and "Video Communication for Compressing and Expanding Digital Color Video Signal". System and method "US Pat.
It is summarized in the issue.

In the prior art, a three-step block coding system for efficiently coding TV signals consists of sampling, transforming and quantifying. The conversion step using a general discrete cosine conversion method is a lumped counting method consisting of matrix multiplication. In particular, the centralization of this counting method, which requires real-time coding, further complicates the actual coding system.

In order to maintain the planar resolution and high frequency components of a TV signal, it is usually necessary to sample at a rate twice that of the highest frequency component of the frequency. For standard broadcast TV, the sampling rate is about 9 megahertz. This results in about 500 samples per horizontal scan line, about 120,000 samples per TV field and about 240000 samples per TV frame. If each sample is 15 bits, then there are approximately 3.6 million bits per frame, and 108 million bits per second. This bit rate has to be reduced for future applications.

[0005]

SUMMARY OF THE INVENTION An object of the present invention is to compress the amount of information transmitted when a video signal is coded and transmitted through a narrow band transmission line such as an intersystem cable or a telephone line. Method and system.

[0006]

In order to achieve the above object,
The system of the present invention (a) divides an analog-to-digital converted video signal into blocks that are a continuous matrix of a plurality of pixels and a bitmap representing a digital word defining first and second reference values. Block coding means for coding, (b) means for finding and identifying non-redundancy information in the coded block, and coding, and (c) each block corresponding to the immediately preceding frame. Inter-frame comparison means for comparing and determining whether there is no difference or a new block,
(D) means for compressing the block and deleting the bitmap when the first reference value and the second reference value are equal in the block which is determined to be a new block, and (e) the current R,
The G and B colors are compared with the value obtained by multiplying the corresponding color of the previous frame or block by the color of the previous block, and when the two are equal or the difference is small, the current color value is different from the previous color value. And an encoder-compressor comprising a compression means for encoding in the form of.
In addition, the method of the present invention uses a block of a matrix of pixels for digitally converting the analog video signal and processing the pixels in order to transmit the video signal at a low bit rate through a narrow band transmission line. Coding, (b) non-redundancy information is found and specified in the coded block, and coded, and (c) each block is compared with the corresponding block in the immediately preceding frame to find no difference or new It is determined whether the block is a block, and (d) if the first reference value and the second reference value are equal in the block that is determined to be the new block, the block is compressed and the bitmap is deleted. ) Compare the current R, G, B colors with the corresponding color of the previous frame or block multiplied by the color of the previous block to see if they are equal or different. It is characterized in that it codes in the form of a difference with respect to the color value immediately before the current color values when again.

[0007]

The video communication system of the present invention transmits coded and compressed video data over an encoder-compressor and decompressor-decoder link. The encoder-compressor of the present invention does not require complex discrete transform steps, but instead uses a relatively simple set of logic to find, characterize, and efficiently code non-redundant information.

In the present invention, the video signal is digitized and coded into blocks of two types of pixels (pixels), a bit map and a frame map. These blocks and frames are easily checked for their redundancy. The novel encoder-compressor feature of the system of the present invention eliminates both block-to-block and frame-to-frame redundancy, reducing the digital data required to represent a video signal. Long run lengths of redundant information are not transmitted as the information is recovered from the bitmap and frame map as it is decoded and reconstructed.

The encoding process has four major steps after digitizing the analog signal: (1) Block coding of individual blocks (matrix of pixels) (2) In-frame filtering to remove noise and leveling between frames to avoid edge effects (3) To eliminate redundancy Intraframe comparison (4) Compression to eliminate all redundancy within frames and blocks

The decoding process also comprises the following steps. (1) Decompression to obtain run-length data (2) Block decoding to obtain luminance and / or color difference (3) Reconstruction of digital signal frame for reconversion to analog signal

[0011]

DESCRIPTION OF THE PREFERRED EMBODIMENT Referring to the encoder-compressor block diagram of FIG. For example, RGB data output from an NTSC system device such as a video camera, a disc player or a video cassette player is an analog signal,
This signal is converted to a digital signal in an analog-to-digital converter 11 and this digital output representing one line of a video frame is stored in a buffer 12. This line is processed by the block encoder, for example 3
× 3 or 9 pixel block or matrix,
That is, it is divided into nine two-color pixels having a digital intensity value and a bitmap that numerically indicates which of the two colors, A or B, describes the pixel.

The block coding method is such that a finite amount of non-redundant elements within one block, in other words the color A
And that there are two elements specified for color B. The matrix bitmap is made up of one bit per pixel. If the bit in the bitmap is 1, then the pixel is color A,
If 0, the pixel is color B. Thus, each block initially consists of three elements of information, color A, color B and a bitmap.

The size of these information elements depends on (a) the required color accuracy and (b) the number of pixels in the matrix. For example, if a color accuracy of 15 bits is required in a system using the present invention, assuming that the size of the matrix is 3 × 3 elements, (15 × 2) +9 (that is, per pixel) 1 bit) = 39 bits. There are 9 pixels in this matrix, so divide 39 by 9 to get 4 and 1 per pixel.
Equal to / 3 bit compression. Since there are 240 × 500 pixels in one frame of a television, there are 13333 blocks per TV field. The sizes of other matrix blocks can be calculated in the same manner.

Next, referring to FIG. The figure is a three-dimensional representation of a block encoder, in a new 2 color + bitmap encoder, a block of 2 × 4 matrix or pixels and a bit indicating whether the corresponding pixel is color A or color B. The map is shown. The attributes of each pixel are encoded in a digital word represented by a height or digital length consistent with the format shown in FIG. These include the color designation A or B, the difference from the corresponding pixel in the previous frame and previous block. It should be noted that there are 15000 2x4 blocks per TV field since there are 240x500 pixels in one television frame. Digital words are summed to indicate pixel intensity or color values.

Further, since the bitmap encodes a variable word length that specifies which one of the five groups of a map pattern a block belongs to, in accordance with the occurrence distribution probability of a particular map pattern, the bit map is variable. Has a length of.

After block encoding, the block data for each block and frame is sent to the memory 15, where it is delayed by one frame time and exists as the previous frame, and then, in the temporal filter 16 which is a time variable impulse response filter. To be filtered. The inter-frame filter 16 reduces noise that causes blurring of edge portions of image data. These edge effects are caused by the noise voltage, which has an additive effect on the brightness values of the color pixels.

After the filter 16, the current frame memory 17 and the immediately preceding frame memory 15 are checked by a compressor 18 for redundancy between frames and blocks and the difference is calculated. In a frame-to-frame comparison, each coded block is compared to the corresponding block in the immediately preceding frame. Each block is marked with a single bit that defines whether it is a new block or unchanged from the previous block.

This process creates a frame bit map of 1 bit per block. Bitmaps for each frame are distinguished by comparison between frames. This frame bitmap can be further compressed because it contains adjacent regions with and without changes, which results in relatively long "1" runs and relatively long "0" runs. And get. Encoding the run length of the frame bitmap yields 0.5 to 0.8 bits per block.

The following table shows the format for compressing block coded information. Total number of bits Color A: Difference (previous frame) ××××× 00 7 Difference (previous block) ×××××× 10 7 Color value ××××× ××××× ×××××× 1 16 Color B: No color B, no bitmap 11 2 difference (immediately preceding frame) ××××× 00 7 difference (immediately preceding block) ×××××× 10 7 color value ××××× ×××× × × × × × × 1 16 Bitmap: Group A × × × 0 4 Group B × × × × 01 6 Group C × × × × × 011 8 Group D × × × × × 0111 9 Group E × × × × × 01111 10 Group F × × × × × × × × 11111 13

The information describing the blocks marked new in compressor 18 of FIG. 1 is further compressed. The information of each of these blocks is a color A, a color B and a bitmap. In the first step, it is determined whether color B = color A. If so, there is no need to code color B anymore, and the fact that the two colors are equal means that the bitmap is all 0s.
There is no need to encode the bitmap since it means that.

The next step is to determine if the coded color value meets the following criteria: {(Rc + Gc + Bc) / (Rp + Gp + Bp)} × Rp = Rc {(Rc + Gc + Bc) / (Rp + Gp + Bp)} × Gp = Gc {(Rc + Gc + Bc) / (Rp + Gp + Bp)} × Bp = Bc Here, Rc and Gc are now and here. The red, green and blue values, Rp, Gp and Bp are the previous red, green and blue values.

These calculations are based on the color values Rp, Gp and B from the previous frame and the previous block.
is done using p. If the color meets the above criteria and the difference (Rc + Gc + Bc)-(Rp + Gp + Bp) is small, the current color value is coded as the difference to the color value of the previous frame or block.

According to the format table above, the bit length coding scheme for bitmaps is based on the "expected frequency of occurrence" distribution. For example, a particular bitmap pattern encoded in group A occurs more frequently than a bitmap pattern encoded in group B. The pattern of group B occurs more frequently than the pattern of group C. Thus, the average number of coded bits representing the bitmap is less than the number of bits in the bitmap.

The minimum number of bits required to code a single block is Color A Difference (7 bits) + No Color B, No Bitmap (2 bits) = 9 bits. The maximum number of bits required to code a single block is 16 + 16 + 13 = 45 bits. According to the actual experimental results using a wide range of video images, 2
On average about 20 bits per block for both × 4 and 3 × 3 block sizes and about 2.3 per pixel.
Met.

When the correlation between frames is high, the block correlation with respect to the immediately preceding frame is high and the number of coded blocks is small. When the frame-to-frame correlation is low, the block correlation to the previous block (within the frame) is high, and the number of coded bits per frame or video is at most 200,000 (densely correlated low-correlation frames) to less. Over the range of 25000 (relatively coarse detail with high intra-frame correlation).

With reference to FIG. 1, the digital video color signal sent from the compressor 18 is composed of the minimum number of bits required to faithfully reproduce the RGB input signal when transmitted to the host bus 19. It The host bus 19 is
A hard wire that couples to other equipment in a video control system, such as in an aircraft or missile, or when transmitting an image using a telephone line, where the high frequency components of the transmitted image are The bit rate must be low to prevent loss.

FIG. 3 is a series of diagrams showing the concept of the novel block coding procedure. The block coding scheme is described assuming that there is a finite amount of non-redundant elements in the block, namely two elements designated color A and color B. In addition, the matrix bitmap is made up of one bit per pixel. If the bit in the bitmap is 1, then the pixel is color A, and if it is 0, then the pixel is color B. Thus, each block initially consists of three elements of information: color A, color B and a bitmap.

The size of these information elements depends on (1) the required color accuracy (2) the number of pixels in the matrix. If 15-bit color accuracy is required and the matrix size is 3x3, (1
5 × 2) + 9 = 39 bits. Since there are 9 pixels in this example, 39/9 = 4 and 1/3 bit per pixel.

The assumption that there is at most two non-redundant elements in a block is verified by examining the maximum expected frequency component in that block. This is easily understood by examining the impulse response on a single axis.
For example, three consecutive samples (a, b,
Examining the three cases with c), FIG.
In case 1 and case 2,
In Case 3, this premise does not hold. However, if the sample rate is 2 × highest frequency component, the opposite is true in case 3. Exactly the same analysis would be performed if samples of three layers were examined.

Referring to FIG. 4, four in a single block
It is assumed that four different colors are contiguous within the block, as shown with the three color elements. Since only 2 colors are allowed in the premise,
Obviously, two of the four colors are distorted. Two
Although the methodology of characterizing the two colors certainly yields the least distortion in the luminance domain, the human eye is not well known to be sensitive to high frequency color information as in the case of FIG. So the visible distortion caused by the block coding algorithm is minimal.

Therefore, a block coding algorithm that only allows the presence of two colors in any block is based on the assumption that any distortion caused by this assumption is acceptable. Various block sizes (2x2, 3x2, 4x2, 2x3, 3x3, 4x3,
2x4, 3x4 and 4x4) are modeled into the actual image coded and decoded to observe the visible distortion. Block sizes smaller than this result in low distortion but poor coding efficiency. Compared to the above values for a 3x3 block with 4 and 1/3 bits per pixel, a 2x2 block has ((2x15) +4) / 4 = 8 and 1/2 bits per pixel.

If the block coding scheme is the only basic element used, the trade-off between efficiency and distortion is made by its own advantages. However, due to the coding techniques between consecutive frames and within frames, small block sizes are almost as efficient as larger block sizes. After the inter-frame and intra-frame redundancy has been characterized, the difference in coding efficiency with block size is negligible.

Explaining the interframe filter 16 of FIG. 1, the function and novelty of this filter are as follows. A block of coded information (colors A and B) passes through this filter 16 after block coding and before interframe comparison. This filter is perfectly described as a time variable, impulse response filter. The purpose of this filter 16 is to reduce noise.

The algorithm of this filter is as follows. That is, if the absolute value of (new sample-stored value) <threshold value, then block color = (coefficient X × new sample) + (coefficient Y × stored value) where coefficient X + coefficient Y =
1 Coefficient X / coefficient Y is time variable Coefficient X / coefficient Y = 1 when time 0 0 <coefficient X / coefficient Y <1 accumulated value of frame N = block color of frame N−1 For example, block color = new sample

To actually execute this algorithm,
Use the accumulated count value for each color in each block. The total number of counts accumulated is 2 × number of blocks. When each block is newly coded, its color value is compared to the stored value. If the difference exceeds the threshold, the new value is stored as is and the count value is reset to zero. If the difference does not exceed the threshold, the count value or number of frames is used to index the table to find the coefficient X and the coefficient Y. These coefficients are then multiplied by the new sample and the stored value and the products are added and stored as the filter value. Finally, if the count is less than the preset maximum value, the count is incremented.

As described above, each color in each block has its own time variable coefficient. Furthermore, when the new block color exceeds the threshold value, it immediately becomes the current value and the time variable coefficient starts from 0 again, so that the edge portion is not blurred.

FIG. 5 shows a functional flow diagram of the decompressor-decoder. This is basically the reverse process of the encoder-compressor. The decompressor 21 reconstructs each frame from the frame bitmap and stores it in the frame memory 22, and the block bitmap is then used to look up the index table of the block decoder 23 to decode the color values. This processing is done on a line basis in the reconstructor 25 based on the results stored in the line buffer 24 to apply the error correction algorithm. The output of the reconstructor is digital to analog converted to reproduce an RGB color continuous video frame signal.

Further compression of the data is done in compressor 18, where the current color is compared to the previous block or previous frame. The scheme does not code the current individual RGB color changes, but instead codes the difference between the sum of the current and previous RGB values.
For example, the RGB values are compared and summed, as shown below. If the ratio of this sum multiplied by the previous value is less than or equal to the set tolerance, the current RGB values are not coded, only the difference in the sum is coded, further compressing the amount of data to be transmitted.

Immediately-Previous Current Total Ratio x Immediately (Calculated Color) R 15 20 18.75 G 20 25 25 25 B 25 30 31.25 Total 60 75 The process takes the ratio of totals (75/60 = 1 .25) Multiply the individual RGB colors by the ratio of the totals Compare the calculated colors to the actual colors Encode the difference (75-60) or 15 because it is an acceptable error

This compression technique is based on empirical data that for 90% of the time the sum ratio equals the ratio of the individual color changes. If the calculated color is outside the tolerance, the current individual RGB color value is coded.

While particular embodiments of the present invention have been illustrated and described, modifications and variations are readily apparent to those of ordinary skill in the art and, therefore, the appended claims are intended to cover such modifications and equivalents. It will be understood.

[0042]

As described above, the method and system for encoding and compressing the video signal of the present invention is transmitted when the video signal is encoded and transmitted through the intersystem cable or telephone line. It has the unique effect of being able to compress the amount of information and reduce the bit rate to a value allowed by those lines.

[Brief description of drawings]

FIG. 1 is a block diagram of an encoder-compressor.

FIG. 2 is a block encoder diagram illustrating a method and bitmap encoding of color pixels of a digitized television signal.

FIG. 3 is an explanatory diagram showing non-redundancy elements in a block.

FIG. 4 is an explanatory diagram assuming that four different colors intersect each other in a block.

FIG. 5 is a block diagram of a decoder-decompressor.

[Explanation of symbols]

11 analog-digital converter 12 line buffer 14 block encoder 15 frame memory 16 inter-frame filter 17 frame memory 18 compressor 19 host bus 21 expander 22 frame memory 23 block decoder 24 line buffer 25 reconstructor 26 digital-analog converter

─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] October 27, 1992

[Procedure Amendment 1]

[Document name to be amended] Statement

[Name of item to be corrected] Claim 7

[Correction method] Change

[Correction content]

[Procedure Amendment 2]

[Document name to be amended] Statement

[Name of item to be corrected] Claim 9

[Correction method] Added

[Correction content]

9. Identifying the bit of the block code to identify the coding type of the block coded data,
Extract color information and bitmap data from the block code, convert the extracted color information and bitmap data into numerical values using a conversion table, and apply this color numerical value to individual pixels in each block according to the numerical bitmap data. And send the resulting digital pixel data to a digital-to-analog converter for conversion into a real-time analog video signal.
A method for decompressing and decoding digital data encoded and compressed according to. ─────────────────────────────────────────────────── ───

[Procedure amendment]

[Submission date] February 18, 1993

[Procedure Amendment 1]

[Document name to be amended] Statement

[Name of item to be amended] Claims

[Correction method] Added

[Correction content]

Claims (8)

[Claims] 1. A code for converting an analog-to-digital converted video signal into blocks which are a continuous matrix of a plurality of pixels and a bitmap representing a digital word defining first and second reference values. Block coding means for coding, (b) means for finding and identifying non-redundancy information in the coded block, and coding, and (c) comparing each block with the corresponding block of the immediately preceding frame. If the first reference value and the second reference value are the same in the inter-frame comparison means that determines whether there is no difference or is a new block, and (d) the block that is determined to be the new block Means for compressing blocks and deleting bitmaps, and (e) the current R, G, B colors and the corresponding color of the previous frame or block And a compression means for comparing the current color value in the form of a difference with respect to the immediately previous color value when they are equal or the difference is small. System for encoding and compressing featured video signals. 2. The encoder-compressor further stores each reference value of each block in the count memory by 2 × block number and takes the difference between the current value and this stored value, and this difference is set. If the difference is smaller than the set threshold value, the new block value is used as it is and the time variable coefficient is started from 0. If the difference is smaller than the set threshold value, the coefficients X and Y are searched from the table. An interframe filter means for multiplying the new value and the stored value, adding them together and storing as a filtered value, and increasing when the count value is smaller than a preset maximum value is provided. A system for encoding and compressing a video signal according to claim 1. 3. The encoder-compressor further creates a frame bitmap by comparing each block with the corresponding block in the immediately preceding frame to determine if there is a change and to mark it with a 1-bit mark. For coding and compressing a video signal according to claim 1, characterized in that it comprises inter-frame comparison means for comparing the frame bitmaps of the current and immediately preceding frames to determine the run lengths of the same and different blocks. System of. 4. In order to transmit the video signal at a low bit rate through a narrow band transmission line, (a) the analog video signal is digitally converted and coded into a block of a matrix of pixels for processing the pixel, (B)
Non-redundancy information is found and specified in the coded block, coded, and (c) each block is compared with the corresponding block of the immediately preceding frame to determine whether there is no difference or a new block. Then, (d) if the first reference value and the second reference value are the same in the block that is the new block, the block is compressed and the bitmap is deleted, and (e) the current R and G , B and the corresponding color of the previous frame or block multiplied by the color of the previous block are compared, and when the two are equal or the difference is small, the current color value is subtracted from the previous color value. Method for encoding and compressing a video signal characterized by encoding in form. 5. In a video transmission system, each reference value of each block of block coded video information is counted and stored, and the newly coded block is compared with this stored value. If the difference is larger than the set threshold value, the new block value is directly stored as the current block value and the time variable coefficient is set to 0.
If the difference is smaller than the set threshold value, the coefficients X and Y are searched from the table by the frame value, multiplied by the new value and the accumulated value, respectively, and then added as a filtered value. An inter-frame filtering method in which noise is reduced by accumulating and then increasing when the count value is smaller than a preset maximum value. 6. The video data is sampled and digitized, and this is formed into a block of a matrix consisting of two kinds of pixels of A or B provided with luminance value information, and each pixel in the block has a bit for A. 0 again
A method of block coding video data, characterized by creating a map of the positions of pixels A and B within a block by appending bit 1 to B respectively. 7. A block decoding of a matrix consisting of pixels having two kinds of elements and brightness and a bitmap of the two kinds of elements, and reproducing the alternation of the two kinds of elements for each frame with reference to an index. A decompressor-decoder for a video transmission system, characterized in that each frame is reconstructed by a frame bitmap to decompress video data. 8. (a) A digital television signal is coded into a block of pixels having one of two types of reference values A and B and a luminance value, a bit map of each block is coded, and both are coded. A block coding means for combining and outputting signals; (b) means for accumulating a block of at least one previous frame connected to the output of the block coding means; and (c) a block coding means and accumulating means. Interframe comparison means connected and comparing each current block with the block at the corresponding position in the previous frame to specify that the block is a new block, and specifying the current block as new and outputting
(D) means for comparing the new block input from the interframe comparing means with the immediately preceding block and deleting the bitmap when the reference value is equal to the value of the immediately preceding block,
(E) Compare the new block with the previous block and the previous frame to determine if the difference is less than or equal to the allowable limit, and to set the current value in the form of the difference to the sum of the current and previous RGB color values. Video communication comprising means for outputting a coded and compressed digital signal, and (d) means for receiving, expanding, decoding, reconstructing and D / A converting the compressed digital output. system.