KR100462855B1 - Differential Order Video encoding System - Google Patents

Abstract

TV communication methods and systems suitable for high definition narrow bandwidth equipment, such as dedicated broadcast systems and video links, that use a block coding process to create a system that does not waste bandwidth caused by transmitting redundancy from multiple sources that do not contribute to picture quality. Is released. Video according to the need for differential order video encoding method (DOVE) can carry 50% of the bandwidth of a general purpose 4.5MHz residual sideband suppression TV system and provide true high quality video by real time TV encoding 14 Provide coding techniques. An efficient coding algorithm based on finding and removing multi-source redundancy characterizes redundancies and generates in real time an efficient code set that encodes only the difference information, eliminating the need for a full 4.5 MHz bandwidth.

Description

Differential Order Video Encoding System

TECHNICAL FIELD The present invention relates to television means and methods, and more particularly, to a television system that reduces bandwidth waste by conserving bandwidth caused by transmitting multiple redundancies that do not contribute to pixel presentation quality. In particular, the present invention relates to coding systems for transmitting only non-repetitive pixel data. There is a demand for high definition TV at relatively narrow bandwidth.

TV signal compression by the block coding method is the subject of known patents of the inventors of the present invention. These known patents relate to specific algorithms for linear coding of certain kinds of redundant data, not to adaptive algorithms as in the present invention.

In the sampled digital video, there is substantially redundancy in order to maintain the resolution needed to sample at a rate at least twice the highest frequency component contained in the analog video stream. In the case of NTSC video, this corresponds to at least 9 MHz (2 x 4.5 MHz), but there is some redundancy in the sample since the actual video has some low frequency components.

Other redundancies are identified by oversampling the depth of each sample. For example, if 8 bits per color are digitized into three colors (e.g., R, G, B, or Y, U, V), the sample You can get 24 bits per second. Each sample can represent one of up to 16.7 million colors. If we sample at 512 x 480, we can get 245,760 pixels. Although all 245,760 pixels can be represented in different colors, for any given video picture, only 18 bits are needed to represent 245,760 colors. In fact, since one picture consists of objects of relatively constant color, the actual number of colors within a given video frame is much less than 245,760.

Another form of redundancy in video is interframe. In a series of individual video frames, it is reasonable to predict that the differences between the frames appear relatively small. Using interframe redundancy can result in an additional compression rate of about 600%, but interframe redundancy does not always exist. Depending on the actual individual video, the interframe redundancy may be very high or zero.

1 is a flow diagram of the functional aspects of the essential process elements of the differential order video encoding (DOVE) coding method of the present invention.

The present invention relates to a block coding method for TV communication, which is particularly suitable for narrowband high definition devices such as picture transmission dedicated systems and video links. This block coding method provides a system that does not waste bandwidth caused by transmitting redundancy from multiple sources that do not contribute to the quality of picture presentation. The present invention utilizes a differential order video encoding (DOVE) technique to generate true high quality video with real-time TV encoding for transmission at half the bandwidth of a 4.5 MHz residual sideband suppression TV system. An efficient coding algorithm based on the concept of detecting and eliminating multiple source redundancy characterizes redundancy, and then generates an efficient code set for encoding only different information in real time, eliminating the need for a full 4.5 MHz bandwidth.

The purpose of differential order video encoding (DOVE) is firstly to provide nearly uniform interframe coding for various pictures, and secondly, to account for the wide variability of interframe redundancy, Provide interframe coding that provides good results at data rates; thirdly, always provide high-quality video that conforms to TV broadcast standards; and fourthly, video that can be easily implemented with microcircuit technology. It is to provide a coding method.

A passive order video encoding (DOVE) system according to the present invention utilizes low frequency color components by subsampling U and V for Y samples using YUV color coordinates.

Although known in the art, producer standards and sign definitions for color video are described below. In other words, see H. W. Sams and Co., Inc., ITT Corp., "Reference Data for Radio Engineers," 5th edition, 28-33.

The producer standards are as follows:

E _M = 0.30E _R + 0.59E _G + 0.11E _B

E _I = 0.06E _R -0.28E _G -0.32E _B

E _Q = 0.21 EPVR-0.52 E _G + 0.31 E _B

U is color coordinate E _Q , V is color coordinate E _I , and Y is color coordinate E _M.

dU = pixel U-previous pixel U

dV = pixel V-previous pixel V

dY = pixel Y-previous pixel Y

dx = dU or dV

YO = sum of all pixel values in a block divided by the amount of pixels in a block

locnt = amount of pixels in a block <YO

hicnt = amount of pixels in a block => YO

cpt = sum of all "quantized" pixels in a block divided by the amount of pixels in a block

abs (dx) = dx rounded

sgn (dx) = the sign of dx, i.e. + or-, or 0

7-bit RGB values can be converted to 7-bit YUV values using the following integer arithmetic algorithm based on the world standard RGB ratio. In other words,

Y = ((38 * r) + (75 * g) + (15 * b)) / 128

U = 64 + (((-22 * r) + (-42 * g) + (64 * b)) / 128)

V = 64 + (((64 * r) + (-54 * g) + (-10 * b)) / 128)

The purpose of the +64 offsets for U and V is to allow them to be treated and treated as positive integers. This fixed offset is removed from the algorithm for converting back to an RGB video display as shown below.

red = Y + (178 * (V-64) / 128)

grn = Y-(89 * (V-64) / 128)-(43 * (U-64) / 128)

blu = Y + (222 * U-64) / 128)

The first step of the DOVE coding technique is optimally represented by the nonlinear difference quantizer 11. If the information is regarded as a "difference forming a difference", the concept of difference quantification will be easier to understand. Individual pixels are arranged in blocks. For example, each major block, such as block (6 × 6), is further arranged into four minor blocks (3 × 3). The U and V values of the major block and the four minor blocks can be obtained by dividing the sums of all U and V of the 3x3 block by 9 and the sum of the U and V values of the four minor blocks by 4. To simplify the implementation, dividing by 9 is approximated by integer operations using l / l6 + l / 32 + l / 64. Thus, we get the values of U and V subsampled for Y at a ratio of 6: 1: 1 and 3: 1: 1. These values are quantified using a differential quantizer similar to the quantizer used for the Y component. Models of the U and V quantizers are shown below, where dx means dU or dV.

if abs (dX)> 0 and abs (dX) <2, then dX <--- 1 * (sgn (dX))

if abs (dX)> 1 and abs (dX) <5, then dX <--- 3 * (sgn (dX))

if abs (dX)> 4 and abs (dX) <10, then dX <--- 7 * (sgn (dX))

if abs (dX)> 9 and abs (dX) <17, then dX <--- 13 * (sgn (dX))

if abs (dX)> 16 and abs (dX) <32, then dX <--- 24 * (sgn (dX))

if abs (dX)> 31 and abs (dX) <47, then dX <--- 39 * (sgn (dX))

if abs (dX)> 46 then dX <--- 54 * (sgn (dX))

Test 12 is performed for overflow or underflow in quantizer 11, where if dX + before X> 127 or <0, then dX is the next lower quantification value. And the test is made again. 6 x 6 UV differs from the preceding (horizontal) block and these differences are quantified. Four individual small blocks differ from large blocks. After quantification, if the difference in one of the four small blocks exceeds the limit, a 3x3 block is encoded with the large block (14). Each 6 × 6 block has a single bit associated with it to indicate whether the encoding is for a single large block or for the large block plus one or more small blocks. Thus, the subsampling ratio is a value that varies dynamically between 6: 1: 1 and 3: 1: 1, depending on the setting of the limit for the actual change of U and V. Quantified U and V values are encoded for large blocks and some small blocks, and the values of the small blocks differ from the large block by an adjustable error (UVerr). Some actual encoding results look like this:

** includes overhead for the beginning row and coding of 6: 1: 1 and 3: 1: 1 block characterizations.

For the two sample pixels selected, the UV components are reduced to 0.243 and 0.148 at 14 bits for ratios of 58: 1 and 94: 1, respectively. CCIR specifies 4: 2: 2 UV subsampling for top quality studio work, while NTSC uses 4: 1: 1. By using 3: 1: 1, DOVE is more accurate than NTSC. However, it is not as accurate as top-quality studio video.

Common coding for UV elements is as follows:

With only limited additional complexity, the feature bits for each 6x6 block can be compressed in one row of blocks. That is, it can be compressed by arranging these bits in one stream at the front of each row of blocks. Compression occurs by run-length coding 15 this stream of bits. This is because the probability that the feature bit is "0" (multiple redundant "0") is high.

An algorithm for execution length coding of these bits is described below. The following subroutine has feature bits in one memory element, and the size of its UVbuf is equal to the total number of 6x6 blocks in a row (equivalent to 85 for 510 pixels / H). The resulting run length (1 or 0) is stored in memory LGTHbuf.

rnlgth: 'sub-routine for run length

For run lengths 0s and 1s

lgth = 0: flip = 0 '1st bit is always 0

           That is, there is no exception in the first bit

addrs = 0

for lup = 0 to 84

if UVbuf (lup) = flip then go to rnlg2 'If the above bits are the same, continue

if flip = 0 then flip = 1 'otherwise flip bits

LGTHbuf (addrs) = lgth 'Save execution length

addrs <-increments addrs + 1 'address

lgth = 0 'and reset the length to 0

rnlg2:

lgth <-lgth + 1

next lup

'To get the last length value

LGTHbuf (addrs) = lgth

RETURN

The individual run lengths are coded as follows:

For the purpose of encoding Y information, DOVE uses 3 x 3 blocks, and the Y components are encoded using six techniques.

(1) Average Y (YO) in 3 x 3 blocks

(2) low-amplitude (small block)

(3) medium to low amplitude (medium to small block)

(4) medium to high amplitude (medium to large block)

(5) high-amplitude (large block)

(6) Map Y1 and Y2 based on allowable error

YO is the sum of all the Y's in a 3x3 block divided by 9. To simplify the implementation, dividing by 9 is approximated by integer arithmetic using 1/16 + 1/32 + 1/64. The change, ie difference, of YO is quantified using the YO of the previous block (the first YO in each row is the absolute value YO) (16).

The small blocks are blocks characterized in that the quantized differential Y for all nine pixels is within +/- 1 with respect to quantized YO. Small to medium blocks have a quantized differential within the range +/- 5. Mid-large blocks have all 9 quantified differences within the +/- 26 range and large blocks have all 9 quantified differences within the full range.

By simply subtracting the luminance of adjacent pixels, such as dY = (Yn-Yn-1), the number of bits required to represent dY is doubled compared to Y. Because they all have full-scale values of +/- instead of positive integers. However, redundancy increases substantially because redundancy is much greater at dY than at Y. This is because the change or difference between pixels is so small that it is predictable. For example, the absolute value for Y can be distributed very randomly, but it can show 0, +/- 1 and +/- 2 with a very high probability from the subtraction.

The DOVE method takes advantage of redundancy generated by adjacent pixel differences. Moreover, to further increase redundancy, and to reduce the number of bits needed to represent dY, DOVE quantifies the difference values represented by dY (16). dYout = dYin | Quantification |.

By considering the nature of video information, the concept of quantifying differences can be understood. The human eye is very close to nonlinear, and is sensitive to and detects small differences in absolute brightness between adjacent elements. On the other hand, it is less sensitive to absolute large differences. The imaging device used for a TV camera is a linear device. To get closer to the human eye, TV cameras use a nonlinear correction called "gamma correction", which provides much higher accuracy and sensitivity for lower levels than higher levels of brightness.

Based on an understanding of the human eye mechanism and various experimental results and empirical products, a set of values has been developed that are used to quantify the difference in luminance between adjacent pixels. This quantification value is shown below.

Input values for this quantizer are 0 to +/- 127.

DY QUANTIZER FOR PIXEL Delta Ys AND Delta Block YOs for Pixel Delta Y and Delta Block YO

if dY = 0 then dY <--- 1

if abs (dY) <3 then dY <--- 1 * (sgn (dY))

if abs (dY)> 2 AND abs (dY) <8 then dY <--- 5 * (sgn (dY))

if abs (dY)> 7 AND abs (dY) <19 then dY <--- 13 * (sgn (dY))

if abs (dY)> 18 AND abs (dY) <34 then dY <--- 26 * (sgn (dY))

if abs (dY)> 33 AND abs (dY) <51 then dY <--- 42 * (sgn (dY))

if abs (dY)> 50 AND abs (dY) <68 then dY <--- 59 * (sgn (dY))

if dY> 67 then dY <--- 76 * (sgn (dY))

As pointed out in Fig. 1 (18), the quantified property is checked for overflow and underflow. If the value before quantified dY + is greater than 127 or less than 0, then the next lower quantification value is used and is checked again. There are 14 individual values from quantizer 16 (ie, +/- 1, +/- 5, +/- 13, +/- 26, +/- 42, +/- 59, and + / -76), only 4 bits are required to represent all values for dY. However, because there is a predictable frequency of occurrence for certain values for some encodings, variable bit-length codes are used. As will be explained later, the pixels are allocated in groups or blocks of 3x3 pixels (19) to encode Y degree.

A small block in which all nine pixels have a dY of +/− 1 is coded with 3 bits (12 bits total) for 1 bit plus block ID per pixel. As a result, 1.33 bits / pixel are obtained for a compression of 5.25: 1 as follows.

dY Small-block dY code

-1 1

 1 0

A small-medium block, characterized in that all nine pixels have a dY in the range of +/- 5, is coded with 3 bits (21 bits total) for a 2-bit plus block ID on a pixel, as follows: 3: 1 From the compression ratio of 2.33 bits / pixel is derived.

dY Small to Small Block dY Code

-5 11

-1 10

 1 00

 5 01

A medium-to-large block, characterized in that all nine pixels have dYs in the range of +/- 26, is coded with 3 bits per block ID plus 3 bits per pixel (total 30 bits), as follows. The result is 3.33 bits / pixel at a compression ratio of 1 :.

dY Medium to large block dY code

-26 100

-13 101

 -5 110

 -1 111

  1 000

  5 001

 13 001

 26 011

For large blocks characterized in that the dY for one or more pixels in the block is greater than +/− 26, the coding and substantial statistics of dY for the pixels are presented below. These blocks use variable bit length codes to exploit the expected probability of occurrence frequency for various dY's. The two test pictures shown represent high density and low density pictures.

Quantifying the differential and variable bit-length coding can save nearly 50% in the original Y data, and save about 13% from the direct forward coding of dYs. Of course, the actual coding result will be different for each picture depending on the statistical distribution. Code assignments were chosen to obtain uniform coding results over a wide range of expected distributions of picture density. From the theoretical point of view and by examining the various individual frames of the various ultra-high quality videos, the following facts were determined: That is, it was determined that quantifying the luminance difference between adjacent pixels having the above values is that the difference between the original and the quantized picture is negligibly small. While this quantification can cause distortion to some values, quantizer 16 is designed to keep the discernable distortion to a minimum.

Some examples for quantifying the differences are given below.

Example 1: typical video tilt

Example 2: Impulse Response (63% Full Scale)

Example 3: Impulse Response (100% Full Scale)

Because the bandwidth of the signals is limited, so the rise times are limited, the signal cannot reach zero to full scale in one sample time. The maximum quantization difference corresponding to 76 was selected based on the theoretical maximum slew rate of the signal limited to 4.5 MHz and sampled at 2 × 4.5 MHz. If the rate of change of the input signal is not limited to 4.5 MHz, the result of quantification to a maximum value of 76 will limit the rise time as if the signal were limited to the effective bandwidth of the 1/2 sampling rate.

In the mapping of blocks based on tolerance error, steps 20 and 21 of the coding process provide an important additional level of compression of the Y low. This first divides a 3x3 block by two luminous intensity values Yl, Y2. Yl is the sum of all the pixels in the block having a value greater than Y0 divided by the amount of these pixels, and Y2 is the sum of all the pixels in the block having a value less than Y0 divided by the amount of these pixels.

In both cases the divisor is always an integer in the range of 1 to 8 by the following logic.

if locnt> 0 then Y2 = Y2 / locnt

if hicnt> 0 then Y1 = Y1 / hicnt

if hicnt = 0 or locnt = 0 then Y1 = cpt: Y2 = cpt

Where locnt is the count of pixels less than or equal to YO and hicnt is the count of pixels greater than YO. Because of the microscopic changes due to quantification, discrepancies can occur, such as when nine pixels are larger than YO. cpt is the average of all nine quantified pixels.

If hicnt or locnt is 0, the count of 9 is never used because both Yl and Y2 are forced to equal cpt. Thus, the actual possible count used as the divisor is in the range of 1 to 8. By subtracting 1 from the actual count, the code for the count is 0 to 7, thus only 3 bits. The maximum sum of the eight pixels is 127 × 8 = 1,016, which is a number corresponding to 10 bits. However, since the accuracy of Yl and Y2 is not an important issue, this can be simplified and the Y value used to obtain the sum of the sums of Yl and Y2 of each pixel divided by two is truncated. Therefore, the maximum sum value is 63x8 = 504, which corresponds to a 9-bit value, with 9 bits being the dividend and 3 bits being the divisor.

The required division can be performed in a read-only memory lookup table. The 9-bit dividend plus the 3-bit divider forms a 12-bit address for read-only memory (ROM). The 7-bit data from this is the quotient (4096 x 7 = 28,672 bit ROM). The Y value is truncated at the end. Therefore, the maximum sum value is 63 × 8 = 504, resulting in 9 bits, followed by a divided number (dividend) of 9 bits and a dividing number (divisor) of 3 bits.

Using this simple integer arithmetic, we get the values for Yl and Y2. This step is also called luminance partitioning. The next step is to generate a 9-bit map, where 1 means that the appeal will be represented by Yl, and 0 means that the pixel will be represented by Y2.

After the map is generated (1 bit per pixel), we have an option of a stage called REMAP. After luminance division and mapping, the block is decoded, and pixel-by-pixel comparison is implemented with the values of the originals in a particular block. If the error generated by the mapping is within acceptable limits, the block is encoded into the mapped block. Otherwise, as indicated above, it may be one of a small block, a small block, a large block, or a large block.

There are two aspects to error tolerance. One is the error in Y caused by the mapping, and the other is that N pixels exceed the error. Each of the three types of blocks (small, medium, and large blocks) goes to the mapping process and is checked for errors (21). Since the small blocks are encoded with only 1 bit per pixel, the mapping process does not provide additional compression. Thus, small blocks are not provided for the mapping process.

Quantifying and encoding the YO difference (23) is implemented by using the YO of the previous block, where the first YO of each row is the absolute value YO, and the quantized difference Y for all nine pixels is quantified. Small blocks are characterized by blocks that fall within a range of +/− 1 for, small blocks have a quantified difference within the range of +/− 5, and large blocks have a total of nine quantified differences within the range of +/− 26. Has Large blocks have nine quantified differences at full scale. The process of encoding 25 the previously unquantized quantified differences is performed, and the process of buffering and organizing data 26 for the output is performed.

Thus, a novel differential order video encoding system (DOVE) has been described that satisfies all purposes and advantages. Various changes, modifications, alterations, other uses, applications, etc. of the present invention will become apparent upon reading this specification in conjunction with the drawings and claims. All such changes and modifications, etc. without departing from the spirit and scope of the invention should be considered to be encompassed by the invention, which is limited only by the claims.

Claims (15)

A digitally coded data compression TV system for realizing high quality with a narrow bandwidth, the system comprising: Means for calculating the Y intensity difference of adjacent pixels and the Y intensity difference of adjacent average blocks, Means for generating non-linear quantification of adjacent pixel and adjacent Y brightness differences, and Variable bit length encoding means of the nonlinear quantized pixels and quantized block Y luminosity Data compression TV system comprising a. The method of claim 1, Luminescence difference quantification means for encoding Y information using 3 × 3 blocks and for encoding Y components using one of the following techniques Data compression TV system, characterized in that it further comprises. a) Average Y (YO) of 3 × 3 blocks b) low amplitude (small block) c) low to medium amplitude (small and medium block) d) used amplitude (medium block) e) high amplitude (large block) The method of claim 1, subtraction means for subtracting luminosity of adjacent pixels such that dY = (Yn-1), said subtraction means for doubling the number of bits required to represent dY relative to Y Data compression TV system, characterized in that it further comprises. The method of claim 1, Means for finding YO by the sum of all the Y's in a 3 × 3 block divided by 9 Wherein, dividing by 9 is approximated by an integer operation using l / l6 + l / 32 + l / 64. A digitally coded data compression TV system for realizing high quality with a narrow bandwidth, the system comprising: Means for calculating the Y brightness difference of adjacent pixels, Means for generating nonlinear quantification of the Y brightness difference of adjacent pixels, Variable bit length encoding means of Y brightness of the nonlinear quantized pixel, and Means for quantifying and encoding the YO difference using the YO of the previous block so that the first YO of each row is an absolute value YO Wherein the small blocks are characterized as blocks in which the quantized difference Y for all nine pixels is within the range of +/- 1 for the YO that is quantified, and the small blocks are the quantified difference within the range of +/- 5 The large block has nine quantified differences in the range of +/- 26, wherein the large block has nine quantified differences in full scale. system. A digitally coded data compression TV system for realizing high quality with a narrow bandwidth, the system comprising: Means for calculating the Y brightness difference of adjacent pixels, Means for generating nonlinear quantification of the Y brightness difference of adjacent pixels, Means for encoding variable bit lengths of Y luminosity of the non-linear quantized pixels, Means for transmitting the coated signal for remote reception, and Means for receiving and decoding the transmitted signal using the following integer arithmetic algorithm Data compression TV system comprising a. red = Y + (178 * (V-64) / l28) grn = Y-(89 * (V-64) / l28)-(43 * (U-64) / l28) blu = Y + (222 * (U-64) / l28) A data compression TV system digitally coded in real time for realizing high quality with a narrow bandwidth, the system comprising: Means for converting RGB analog TV signals to digital YUV coordinates by converting x-bit RGB values to x-bit YUV values according to an integer arithmetic algorithm based on world standard RGB ratios including analog-to-digital conversion, Means for quantifying and encoding the difference between U and V values based on one major 6 × 6 block and four minor 3 × 3 blocks, Means for quantifying the difference in luminance for encoding Y information using 3 × 3 blocks and for encoding Y components using one of the following techniques, a) average Y (YO) of 3 × 3 blocks b) low amplitude (small block) c) low to medium amplitude (small and medium block) d) used amplitude (medium block) e) high amplitude (large block), and f) Yl, Y2 mapped based on allowable error subtraction means for subtracting the luminosity of adjacent pixels such that dY = (Yn-1), said subtraction means for doubling the number of bits required to represent dY relative to Y, A means of finding YO by the sum of all the Y's in a 3x3 block divided by 9, where dividing by 9 approximates an integer operation using l / l6 + l / 32 + l / 64 Finding Means, Means for quantifying and encoding the YO difference using the YO of the previous block so that the first YO of each row is an absolute value YO, where quantified difference Y for all nine pixels is + / Small blocks are characterized as blocks within the range of 1, small blocks have quantified differences within the range of +/- 5, and large blocks have all 9 quantified differences within the range of +/- 26, where Means of YO difference quantification and encoding, wherein a large block has all nine quantified differences in full scale, Means for transmitting the coded signal for remote reception, and Means for receiving and decoding the transmitted signal using the following integer arithmetic algorithm Data compression TV system comprising a. red = Y + (178 * (V-64) / l28) grn = Y-(89 * (V-64) / l28)-(43 * (U-64) / l28) blu = Y + (222 * (U-64) / l28) 8. The data compression TV system according to claim 7, wherein said means for converting an RGB analog TV signal in accordance with said integer arithmetic algorithm comprises an analog-to-digital converter. As a coding method for high definition narrow bandwidth video, Convert the color video signal to a digital YUV value, Perform nonlinear difference quantification to arrange the individual pixels in the major block, wherein each said major block is further arranged into minor blocks, whereby the U and V values for the major block and the minor blocks are minor The sum of all U and V in the block is divided by the number of pixels in the minor block, giving the subsampled U and V values for Y, quantify U and V values according to dx, where dx is dU or dV, if abs (dX)> 0 and abs (dX) <2 then dX <--- 1 * (sgn (dX)) if abs (dX)> 1 and abs (dX) <5 then dX <--- 3 * (sgn (dX)) if abs (dX)> 4 and abs (dX) <10 then dX <--- 7 * (sgn (dX)) if abs (dX)> 9 and abs (dX) <17 then dX <--- 13 * (sgn (dX)) if abs (dX)> 16 and abs (dX) <32 then dX <--- 24 * (sgn (dX)) if abs (dX)> 31 and abs (dX) <47 then dX <--- 39 * (sgn (dX)) if abs (dX)> 46 then dX <--- 54 * (sgn (dX)) -Test for overflow and underflow from quantification, whereby if dx + previous x is greater than 127 or less than 0, dx becomes the next lower quantification value, which is then tested again, Obtaining the difference between the major blocks U and V from the preceding quantification step, quantifying the difference of the minor blocks from the large block, and encoding the minor block together with the major block if the difference of one of the minor blocks exceeds the limit, Allocate one bit associated with the major block to indicate whether the encoding is for a single large block or for the large block and one or more small blocks, Encode the quantized U and Y differences encoded for the major block and minor blocks whose values differ from the major block by an adjustable error (UV (err)), Perform luminance difference quantification to encode Y information using minor blocks and Y components, -Encode using one of the following techniques: a) average Y (YO) of 3 × 3 blocks b) low amplitude (small block) c) low to medium amplitude (small and medium block) d) used amplitude (medium block) e) high amplitude (large block), and f) Yl, Y2 mapped based on allowable error -subtract the luminance of adjacent pixels so that dY = (Yn-1), doubling the number of bits needed to represent dY compared to Y, Find YO by the sum of all the Y's in the minor blocks divided by the number of pixels in the minor block, Quantify and encode the YO difference using the YO of the previous block so that the first YO of each row is an absolute value YO, where the quantized difference Y for all pixels is within +/- 1 range for the quantized YO Small blocks are characterized as blocks, small blocks have quantified differences within the range of +/- 5, and large blocks have all nine quantified differences within the range of +/- 26, where large blocks have quantification Have all the differences in full scale, and Formatting the encoded data by bit packing and organizing the coded data to accommodate the video signal format. And coding the high definition narrow bandwidth video. The method of claim 9, Run length coding the U and V feature bits to remove redundant zeros And further comprising the step of coding a high definition narrow bandwidth video. 10. The method of claim 9, wherein the major block is a 6x6 block and the minor block is a 3x3 block. 10. The method of claim 9, wherein the major block is an 8x8 block and the minor block is a 4x4 block. A coding means for high-quality narrow bandwidth video transmission, the coding means, Means for converting color video signals into digital YUV values, Means for performing nonlinear difference quantification to arrange the individual pixels into major blocks, wherein each said major block is further arranged into minor blocks, whereby U and V values for the major blocks and minor blocks Such non-linear difference quantification means, wherein the sum of all U and V in this minor block is divided by the number of pixels in the minor block to provide subsampled U and V values for Y, Means for quantifying U and V values according to dx by the formula below, where dx is dU or dV, if abs (dX)> 0 and abs (dX) <2 then dX <--- 1 * (sgn (dX)) if abs (dX)> 1 and abs (dX) <5 then dX <--- 3 * (sgn (dX)) if abs (dX)> 4 and abs (dX) <10 then dX <--- 7 * (sgn (dX)) if abs (dX)> 9 and abs (dX) <17 then dX <--- 13 * (sgn (dX)) if abs (dX)> 16 and abs (dX) <32 then dX <--- 24 * (sgn (dX)) if abs (dX)> 31 and abs (dX) <47 then dX <--- 39 * (sgn (dX)) if abs (dX)> 46 then dX <--- 54 * (sgn (dX)) Means for testing for overflow and underflow from said quantification means, whereby if dx + previous x is greater than 127 or less than 0, dx becomes the next lower quantification value, which is then tested again. , These test means, Means for obtaining the difference between major blocks U and V from prior quantization and for quantifying the difference between minor blocks from a large block, and encoding the minor block together with the major block when the difference of one of the minor blocks exceeds a limit. Such means, Means for allocating one bit associated with the major block to indicate whether the encoding is for a single large block or for the large block and one or more small blocks; Means for encoding the quantized U and Y differences encoded for the major block and minor blocks whose values differ from the major block by an adjustable error (UV (err)), Means for performing luminance difference quantification to encode Y information using minor blocks and Y components, Means for encoding using one of the following techniques, a) average Y (YO) of 3 × 3 blocks b) low amplitude (small block) c) low to medium amplitude (small and medium block) d) used amplitude (medium block) e) high amplitude (large block), and f) Yl, Y2 mapped based on allowable error means for doubling the number of bits needed to represent dY relative to Y by subtracting the luminance of adjacent pixels such that dY = (Yn-1), Means for finding YO by dividing the sum of all Y's of minor blocks by the number of pixels in the minor block, Means for quantifying and encoding the YO difference using the YO of the previous block such that the first YO of each row is an absolute value YO, wherein the quantized difference Y for all pixels is in the range of +/- 1 for the YO Small blocks are characterized by blocks within, small blocks have quantified differences within the range of +/- 5, and large blocks have nine quantified differences within the range of +/- 26, where the large blocks Means for quantification and encoding, having all the quantified differences at full scale, and Means for formatting the encoded data by bit packing and organizing the coded data to accommodate the video signal format. And coding means for high-definition narrow bandwidth video transmission. The method of claim 9, By dividing the minor block into two luminous intensity values, Yl and Y2, bit-mapped the blocks based on an error that is acceptable to provide an additional level of compression of Y information, where Yl is every pixel in the block with a value greater than YO. Is the sum of these pixels divided by the amount of these pixels, Y2 is the sum of all the pixels in the block with a value less than or equal to YO divided by the amount of these pixels, and the divisor in both cases is Is an integer ranging from 1 to 8, if locnt> 0 then Y2 = Y2 / locnt if hicnt> 0 then Y1 = Y1 / hicnt if hicnt = O or locnt = O then Y1 = cpt: Y2 = cpt Using simplified integer arithmetic, obtain the values for the average Y, and Y2 for all pixels above Y1, YO, and Generate a 9-bit map, where 1 means that the pixel will be represented by Y1 and 0 means that the appeal will be represented by Y2, And coding a high definition narrow bandwidth video. The method of claim 14, Remapping the processing steps using a lookup table to convert the bitmaps into variable bit length codes based on the expected occurrence probability. Further comprising a step wherein the remapping process is based on assumptions about an acceptable error, where (1) the probability is low and (2) is acceptable to reject some bit maps that are less likely to cause a distinguishable error. Assuming an error, the map takes the form 010 101 101 or 010 010 101 Here, one of these maps occurs only in the case of low amplitude blocks, in which case an illogical map bit pattern is caused by noise, and the lookup table for remapping includes 9 bits with 512 maps, and a variable bit length code The output of the high-quality narrow bandwidth video coding method characterized by representing 84 maps.