MXPA01005570A - Improvement of fine granularity scalability using bit plane coding of transform coefficients - Google Patents

Abstract

A system for efficient bit plane coding of transform coefficient data, such as DCT data used in a video coding system. Decimal values for the transform coefficients, e.g., in a block of several coefficients, are converted to binary values, where each bit occupies a corresponding bit plane, from the most significant bit to the least significant bit. One bit from each coefficient is provided in a common bit plane. A one-bit flag or codeword (such as"0") is used for coding - one or more initial all-zero bit planes, while another one-bit flag (such as"1") is used for designating the first subsequent non-all-zero plane. For the first non-all-zero plane, a reduced coding table is used to provide codewords that follow the one-bit flag. The coding table is reduced in size since it does not require a special"all-zero"codeword. Additionally, the use of a one-bit flag for designating the initial all-zero bit planes reduces the required number of coding bits over prior art schemes that require multi-bit all-zero codewords. An encoder (200) includes a"0"codeword function (242), a"1"codeword function (244), a reduced table (246), and conventional tables (248). A corresponding decoder (400) includes a"0"codeword function (442), a"1"codeword function (444), a reduced table (446), and conventional tables (448).

Description

IMPROVEMENT OF FINE GRANULAR SCALABILITY USING CODING IN BITS PLAN OF TRANSFORMATION COEFFICIENTS BACKGROUND OF THE INVENTION This application claims the benefit of United States Provisional Application No. 60 / 110,882, filed December 4, 1998. The present invention relates to a method and apparatus for efficient bit-plane coding of transformation coefficients, such as Discrete Cosine Transformation (DCT) coefficients. These coefficients can be used in a variety of applications, including digital video encoding and decoding. In particular, there is an improvement in the technique known as Fine Granular Scalability using bit-plane coding (FGSB). The FGSB coding preserves the base layer coding technique intact. For example, the base layer coding technique can be MPEG-2, MPEG-4, or any image / video coding technique based on DCT. In the base layer, the DCT coefficients are encoded using relatively common quantization to obtain low speed bit data. Figure 1 illustrates a prior art apparatus for Fine Granular Scalability using bit plane coding (FGSB). With the FGSB coding a difference (or residue) is obtained between the original integer DCT coefficients and the dequantized DCT coefficients. As shown in the encoder 100, for example, an original block of DCT coefficients is quantized in a quantizer 110, then the quantized coefficients are de-quantized (i.e., quantified in an inverse manner) in an inverse quantizer 120 to obtain the coefficients DCT dequantified. A DCT coefficient difference block is the output information of a subtractor 130 and provided for a bit-plane coding function 140 before it is communicated through a channel, for example, the data can be communicated in a network of broadband communication, such as a satellite or cable television network, or a network of computers, such as the local area network (LAN), the metropolitan area network (MAN), the wide area network (WAN), internet, intranet and the Internet. As mentioned, the base layer DCT coefficient data is conventionally encoded. The bit plane coding function 140 includes one or more tables 145 for encoding the bits in each bit plane. Tables include an "all-zero" code word that is used when a bitmap has all values at zero. Since each of the DCT difference blocks typically has few bit planes (for example, up to four to eight planes in typical applications), fine granularity is achieved at a very low cost of complexity. The number of bit planes is determined by the number of bits needed to encode the larger difference values. Essentially, the DCT difference data in the successive bit plane layers can be used to reduce the quantization error of the DCT coefficient. One or more of the bit plane layers can be recovered by a decoder in accordance with the broadband of the available channel and the processing speed of the decoder. The bit plane layers are recovered by starting from the layer carrying the most significant bits of the difference data DCT, then the layer carries the next most significant bits of the difference data DCT, etc. The FGSB coding can be simplified in the following steps: 1. After performing the base layer coding, which is based on DCT, take the difference between the original DCT coefficients and the dequantized DCT coefficients. Find the number of bit planes required to encode this difference block. 2. Find the maximum number of bit planes from all the difference blocks of a video frame. 3. Encode the maximum number of bit planes at the precise beginning of an improvement of the frame in the layer bit stream. 4. Sequentially encode the bit planes of a frame that starts from the Most Significant Bit level (MSB). 5. When a bit plane is encoded, the 2-D symbols are formed of two components. The first component indicates the number of consecutive zeros (for example, zero cycle length) up to the next "1". The second component is a single-bit flag that indicates if there is any "1" to the left in the current bit plane. The second component is therefore an Indicator of Plane Finish (EOP, for its acronym in English). These 2-D symbols are then encoded by entropy. If there is no "l" s in the whole of the current bit plane, a "all zero" symbol is coded. The following illustrates an example of encoding a particular bit plane using the above approach, designated "prior art 1".

In the first line, the "position" is the bit sequence in the bit plane. For example, for an 8x8 bit plane, there are 64 bits, for example, 0-63. In the second line, the bit value is shown, either a binary 0 or 1. In the third line, the 2-D symbols used by the prior art scheme discussed above are shown. Specifically, at position 0, bit value is "1", and the symbol is (0,0). The first component of the symbol, "0", indicates that the zero cycle length symbols is 0 (ie, there is no zero) before the next "1". The second component of the symbol, "0", which is the EOP indicator, indicates that there is at least one subsequent "1" in the bit plane. Note that the symbols are given in decimal numbers, which are later converted to binary (for example, in position 1, the bit value is "0" and the symbol is (2.0). symbol, "2", indicates that the zero cycle length symbols is 2 (ie, there are zeros in positions 1 and 2.) Thus, the distance to the next "1" is the value of bit 2 (is say, position 3) The second component of the symbol, "0", again indicates that there are additional "l's" in the bit plane (ie, in addition to "1" in position 3.) In position 4, the bit value is "0" and the symbol is (1,1) The first component of the symbol, "1", indicates that the zero cycle length symbols is 1 (ie there is a zero at position 4) In this way, the distance to the next "1" is a bit value "1" (ie position 5.) The second component of the symbol, "1", indicates that there are no additional "I's" in the bit plane (des after position 5). The above FGS coding method is set forth in ISO / IEC JTC1 / SC29 / WG11, MPEG98 / M420, December 1998, "Fine Granularity Scalability Using Bit Plane Coding of DCT Coefficients" With this technique, a bitmap is encoded with All zeros using the "all-zero" symbol, regardless of which bit-plane layer is being encoded The all-zero symbol may need to be coded for more than one bit-plane layer for a given DCT block. disadvantageous due to the size of the all-zero symbol, and to the fact that the size of the encoding table increases.In particular, the size (for example, bit length) of the all-zero symbol is established by the entropy coding ( for example, Huffman coding) performed on the 2-D symbols of the bit plane As it is known, when coding source symbols are not equally likely, it is efficient to use variable length code words. The presence of source symbols is used to select code words in such a way that source symbols are more likely to be assigned to shorter code words. With these restrictions, the length of the all-zero symbol is typically two bits or more. In addition, since there are several thousand DCT 8x8 blocks in one image, (for example, consider an NTSC image of 525x480 pixels) the data generated by the all-zero symbol are significant. Accordingly, it would be desirable to provide a method and apparatus for coding an efficient bit plane that enhances the prior art. The system must reduce the number of bits required to indicate the presence of a bit plane in which all values are zero. The system should avoid the need for multiple multiple "all-zero" symbols in the initial MSB bit plane layers of an individual block having all zeros, thereby reducing the data generated to encode the bit plane. The system must improve coding efficiency by reducing the number of symbols in an MSB-level entropy coding table.

Reducing with this the code length of the remaining symbols in the table. The system must be compatible with coding schemes that improve the multiple encoding tables that are used for the different bit-plane layers. Because the probability of the presence of specific bits is different for the different bit plane layers, the coding efficiency can be optimized by making the coding table for the bit plane layer. This concept is further analyzed in the thesis of the PhD in philosophy entitled "Opt imizat ion of Entropy Coding Efficiency Under Complexity Constraints in Image and Video Compression", (1998), Section 4.3, of Dr. Fan Ling, cataloged in the Electrical Engineering Department of Lehigh University, Pennsylvania, E.U.A. The present invention provides a system having the foregoing and other advantages.

SUMMARY OF THE INVENTION The present invention relates to a method and an apparatus for efficient bit-plane coding of transformation coefficients. A method for the efficient coding of a plurality of bit planes in which the data transformation coefficients are carried out, includes the step of providing a code word having a first state (for example "0") which designates that a bit plane has all the binary zeros and a second state (eg, "1") that designates that a bit plane does not have all the binary zeros. The bit plane of the most significant bit (MSB) is coded with the "0" when the bit plane MSB has all the binary zeros. Proceeding from the bit plane MSB to the bit plane of the least significant bit (LSB), each succeeding bit plane having all the binary zeros is encoded with the "0", if any successive bit planes are present, until A first bit plane is reached that does not have all the binary zeros. The first bit plane is coded with "1" followed by at least one code word obtained from a first entropy coding table according to the bits in the first bit plane. Preferably, the code word is a single bit code word. Importantly, the first entropy coding table does not include a multi-bit code word to encode a bit plane with all binary zeros. In this way, the size of this coding table can be glossed in relation to the prior art schemes, which, in turn, reduces the number of bits for coding. A second conventional entropy coding table is provided to encode one or more planes that follow the first bit plane. This coding table includes a multi-bit code word for encoding a bit plane with all binary zeros and therefore is not reduced in size. The transformation coefficient data may include discrete cosine transformation (DCT) data and / or image data. A corresponding decoding method includes the step of (a) providing a decoding function for a code word having a first state (e.g., "0") that designates that one bit plane has all the binary zeros and a second state (eg. example, "1") that designates that a bit plane does not have all binary zeros. Proceeding from the bit plane MSB to the bit plane of the least significant bit (LSB), the "0" is decoded for each successive bit plane having all the binary zeros, if any successive bit planes are present, until A first bit plane is reached that does not have all the binary zeros. The "1" is decoded for the first bit plane and a first entropy decoding table is used to decode at least one code word following the "l'A A related digital signal and encoding and decoding apparatus are also presented.

BRIEF DESCRIPTION OF THE FIGURES Figure 1 illustrates a prior art apparatus for fine granular scalability coding using bit-plane coding (FGSB). Figure 2 illustrates an apparatus for fine granular scalability coding using bit-plane coding (FGSB) according to the present invention. Figure 3 illustrates a coding method according to the present invention. Figure 4 illustrates a decoder according to the present invention.

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a method and an apparatus for efficient bit-plane coding of transformation coefficients. The present invention is illustrated with reference to the following table, in which the symbols coding the bit plane analyzed in ISO / IEC JTC1 / SC29 / WG11, MPEG98 / M420, discussed above, are designated by "prior art 2" , and the symbols for coding the bit plane of the present invention are designated by the "invention". An example DCT difference block of 8x8 is adopted as follows: Position: 0 1 2 3 4 5 6 7 ... 63 Decimal Value: 11 1 10 3 2 1 0 0 ... 0 Bit Planes MSB-0 0 0 0 0 0 0 0. 0 (all zeros) MSB-1 1 0 1 0 0 0 0. 0 MSB-2 0 0 0 0 0 0 0. 0 (all zeros) MSB-3 1 0 1 1 1 0 0. 0 MSB-4 1 1 0 1 0 1 0. 0 Method of "Prior art 2 'MSB-0 (all-zero symbol), MS B-1 (0,0) (1,1), MS B-2 (all-zero symbol), MSB-3 (0, 0), (1,0), (0,0), (0,1) MS B - 4 (0,0), (0,0), (1,0), (AD) Present invention MSB-0 0 (single flag bit = true), MSB-1 1 (single flag bit = false), (0,0), 1,1), MSB-2 (all-zero), MSB- 3 (0,0), (1,0), (0,0), (0,1) MSB-4 (0,0), (0,0), (1,0), (1,1) The base bit or bit plane layer MSB is designated "MSB-0," the second bit layer layer bits MSB are designated "MSB-1," the third bits of the MSB plane layer are designated " MSB-2, "the fourth bit plane layer bits MSB are designated" MSB-3, "and the fifth bit plane layer bits MSB are designated" MSB-4. " MSB-4 is essentially the bit-plane layer of the least significant bit (LSB). The decimal values of the DCT difference coefficients are given in positions 0-63. The corresponding binary values are provided in the different layers. For example, the decimal value 11 corresponds to the binary value 010112, where the first "0" (which is read from left to right) is the MSB. This "0" is therefore provided in MSB-0. The remaining binary digits are distributed in MSB-1 ("1"), MSB-2 ("0") MSB-3 ("1") and MSB-4 ("1"). Note that all the values 0 of the layer MSB in this example are zeros, which indicate that the decimal values of the coefficients in the current block are all less than fifteen. The five bit plane layers allow decimal values to reach 31. Of course, planes with fewer or more bits can be used. Because different DCT blocks have different numbers of bit planes, the bit planes of the MSB layer have a high probability of being all zero. The present invention takes advantage of this condition. First, the coding used by the prior art technique 2 in the previous example is analyzed. The MSB-0 is encoded using the "all-zero" symbol, which typically requires two or more coded bits. This indicates that the layer is all zeros. For example, you can use a coding table that assigns a code word to a symbol. The number of encoded bits will vary based on the size of the coding table and the entropy coding scheme used. The MSB-1 is encoded using the symbol (0,0), where the first "0" indicates a first zero-cycle length of zero (ie, without zeroes) before the next "1", and the second "0" "indicates that there are additional" l's "on the left in the bit plane after the next" 1". Then, the symbol (1,1) is used where the first "1" indicates a zero cycle length of one (ie, a zero) before the next "1" and the second "1" indicates that there is no " l's "extra to the left in the bit plane after the next" 1". The MSB-2 is encoded using the symbol "all-zero", which indicates that the layer is all zeros. The layer MSB 3 ("1") is coded using the symbol 0, 0 Next, the position 1 '0' is coded using the symbol (1,0), position 3 ("1") is coded using the symbol ( 0,0) and position 4 ("1") is coded using the symbol (0,1). Position 0 of MSB-4 ("1") is coded using the symbol (0,0), position 1 ("1") is coded using the symbol (0,0), position 2 ("0") it is coded using the symbol (1,0) and position 4 ("0") is coded using the symbol (1,1). Accordingly, the prior art technique 2 uses the "all-zero" symbol twice, in MSB-0 and MSB-2. This technique is not optimal since the "all-zero" symbol has a length of two bits or more and the presence of this symbol increases the length of the coding table. By contrast, the present invention allows the use of a reduced coding table, which does not have the all-zero symbol as an entry, for the coding of MSB-0 and MSB-1. Specifically, with the present invention, a single bit flag is introduced to indicate whether the bit plane MSB is all zeros or not. For example, "0" can indicate all zeros and a "1" can indicate not all zeros, or vice versa. However, any time a "1" is encoded, the flag is not coded for all of the following lower-level bit-level planes of the block. In this way, the coding scheme of the invention is applied to the initial MSB layer (layer 0) when it has all the zeros, to any subsequent layers immediately having all the zeros, if these layers are present and to the next subsequent layer that does not have all the zeros. If any additional all-zero bit planes are found (for example, after the bitmap coded with the flag "1"), they are encoded using the "all-zero" symbol of the prior art 2. This is used in MSB-0, the code word or flag (0) to indicate that layer MSB 0 is all zeros. In MSB-1 (1) (0,0) it states that the code word "1" must be followed in the bit stream by the code word for (0,0) and the code word for (1,1), from which two code words are obtained later from the reduced coding table. The total number of bits is reduced in relation to the prior art scheme 2 since a small size coding table is used that does not have the all-zero symbol as an entry. This bit saving can be significant when multiplied with respect to the many transformation blocks. For example, consider the number of DCT blocks of 8x8 in a television image of 720x480. Figure 2 illustrates an apparatus for fine granular scalability coding using bit-plane coding (FGSB) according to the present invention. Elements with similar numbers are used in correspondence with other of the Figures. Here, a revised bit-plane coding function 240 is used, which includes a codeword function 242"0" to encode a code word "0" for an all-zero bit plane as discussed herein. In the previous example, the MSB-0 is coded by this function 242. A code function 244"1" is provided to encode a code word "1" for a candid first bit plane as discussed herein. A reduced table 246 does not include an "all-zero" code word and therefore has a reduced length relative to a conventional table. The reduced length results in a reduced number of coded bits for at least some of the entries in the table. The reduced table is associated with the code word function 244"1" since it is used to encode bits in the bit plane that are associated with the code word "1". For example, in the previous example, MSB-1 is encoded by functions 244 and 246. Specifically, symbols (0,0) and (1,1) in MSB-1 are encoded by function 246. A function 248 of conventional table includes one or more tables with the word code "all-zero". Therefore, these tables have an increased length relative to the reduced table 246 and additional bits will be used to encode a bit plane. One or more conventional tables can be used. If multiple conventional tables are used, each table can be tailored for specific bit planes, as discussed in the background. It is believed that three conventional tables may be suitable for FGSB video applications. In general, the bit probability distribution in the lower bit plane layers becomes more similar for each additional layer, such that the advantage of a conventional coding table is reduced separately. Figure 3 illustrates a coding method according to the present invention. In block 305, the first (ie the most significant) bit plane of a new DCT block is processed. If the initial layer, layer-0 MSB has all the zeros (block 310), the layer is encoded with a codeword "0" (block 315) according to the invention. If the initial layer does not have all zeros, processing continues in block 325.

After block 315, if there are any remaining layers (block 318), a determination is made if any of the immediately following layers also have all zeros (block 320), in which case they are also coded using the "0" symbol (block 315) according to the invention. If there are no remaining layers (block 318), the next DCT block is processed (block 305). In block 325, if the next layer does not have all the zeros, a code word "1" is provided, followed by one or more codewords for the even symbols (e.g., 0.0, 0.1, 1.0). or 1,1) from a reduced table. According to the invention, the reduced table does not have an all-zero symbol, such that the length of the code words for at least some of the symbols is reduced relative to the conventional table. In this way, the total number of bits needed to code a transformation block is reduced. If there are any remaining layers (block 330), the next layer (block 335) is processed using the conventional table (block 340). Figure 4 illustrates a decoder according to the present invention. The decoder 400 includes a function 440 for bit plane decoding, which receives an encoded data stream from a channel. For example, the decoder 400 may be a terminal installed in a network for cable or satellite television to receive digital video data, for example in accordance with the MEPG standard. The function 442 (coding) code word "0" is a counterpart of the function 242 for encoding in the encoder 200 and identifies a coded zero as a designator containing all the zeros of the bit plane. The function 444 (decoder) of the code word "1" is a counterpart of the function 244 for coding in the encoder 200 and identifies a coded one as a designator of a first plane without zero after a plane coded with a code word " 0". Reduced table 446 is a function for decoding that is a counterpart of table 246 in encoder 200, and identifies pairs of symbols corresponding to the code words in the received data stream for the plane that is also identified by a "1"encoded. The conventional tables 448 are decoding functions that are counterparts of the tables 248 in the encoder 200 and identify pairs of symbols corresponding to the code words in the received data stream for the planes that are not identified by a coded "1". Function 440 for encoding the bit plane outputs transformation difference block data for additional conventional processing, the details of which will be apparent to those skilled in the art. Note that the invention is still more efficient for encoding smaller decimal values, where there are several initial MSB layers with all zeros. In general, at least one encoded bit can be saved for each layer that is encoded according to the invention. Typically, most DCT difference values are zero or close to zero, with only slightly larger values whose symbols require the use of the initial MSB layers. However, in any case, the coding efficiency is improved since the number of symbols from the entropy coding table of the MSB level is reduced by a symbol. As a result, the code length of the remaining symbols is also reduced. This can be illustrated by a simplified example by assuming a set of symbols with three symbols,. { A, B, C.}. . Your chances of showing up and the Huffman codes are as follows: Symbol: A B C Probability: 0.4 0.3 0.3 Huffman code: 1 01 00 Code length (bits 1 2 2 If we take a symbol (for example, A) out of the set, the new set will be. { B, C.}. . The previous table is converted to a reduced table Symbol: B C Probability: 0.5 0.5 Huffman code: 0 1 Code length (bit): 1 1 Comparing the two cases, we find that by taking a symbol out of the symbol set, we reduce the code length of all the remaining symbols. This conclusion applies in general to any technique for entropy coding and is not limited to Huffman coding. Note that the total percentage savings in the encoded bits will be less than the number of symbols and the symbol length increases in a table for coding. However, even small savings for each transformation block are significant when multiplied, for example, by the many blocks in an image. Accordingly, it can be seen that the present invention provides a method and apparatus for efficient bit-plane coding that provides a single bit flag or a code word to encode one or more all-zero bit planes and the first flat without any subsequent zero. For the foreground without all zero, a table for reduced entropy coding is used to provide code words that follow the single-bit flag. The coding table is reduced in size since it does not require a special "all-zero" code word. Additionally, the use of a single bit flag to designate the initial all-zero bit planes reduces the required number of bits for coding with respect to the prior art schemes requiring multi-bit all-zero code words . The invention thereby reduces the data generated to encode the bit plane. Although the invention has been described in conjunction with the various specific embodiments, those skilled in the art will appreciate that various adaptations and modifications may be made without departing from the spirit and scope of the invention as set forth in the claims. The invention is suitable for use with any scheme for bit-plane coding and is not limited to bit-plane coding of DCT coefficients. For example, the invention can be used with other techniques for transformation coding, such as Discrete Fourier Transformation, Karhunen-Loeve Transformation, Walsh Hadamard Transformation, wavelength transformation, as well as other known spatial transformations. In addition, any type of entropy coding can be used, such as Huffman coding or arithmetic coding, which uses a floating-point code length. Furthermore, the invention is not limited to coding of MPEG-encoded data or video data, although it can be used, for example, for seismic, vibrational, temperature, pressure and other types of data coding with characteristics 2 - D or higher. Additionally, the invention can be implemented using any known hardware, firmware and / or software techniques.

Claims (22)

1. NOVELTY OF THE INVENTION Having described the present invention, it is considered as a novelty and, therefore, the content of the following CLAIMS is claimed as property: 1. A method for the efficient coding of a plurality of bit planes in which the transformation coefficient data is carried, which comprises the steps of: (a) providing a code word having a first state indicating that a bit plane comprises all binary zeros, and a second state indicating that a plane of bits does not include all binary zeros; (b) encoding the bit plane of the most significant bit (MSB) with the code word in its first state when the bit plane of MSB comprises all binary zeros; (c) proceed from the bit plane MSB to the bit plane of the least significant bit
2. (LSB), coding each successive bit plane comprising all the binary zeros with the code word in its first state, if any successive bit planes are present, until reaching a first bit plane that does not comprise all the binary zeros; and (d) encoding the first bit plane with the code word in its second state, followed by at least one code word obtained from a first table for entropy coding in accordance with the bits in the first bit plane. The method according to claim 1, wherein: the code word having the first and second states is a single bit code word.
3. 3. The method according to claim 1, wherein: the first table for entropy coding does not include a multi-bit code word for encoding a bit plane with all binary zeros.
4. The method according to claim 1, comprising the additional steps of: providing a second table for entropy coding to encode at least one bit plane that follows the first bit plane, where: the second table for entropy coding it includes a multi-bit code word to encode a bit plane with all binary zeros.
5. The method according to claim 1, wherein: the transformation coefficient data comprises discrete cosine transformation (DCT) data.
6. 6. The method according to claim 1, wherein: the transformation coefficient data comprises image data.
7. 7. A method for encoding a plurality of bit planes in which the transformation coefficient data is carried, comprising the steps of: (a) providing a decoding function for a code word having a first state indicating that a bit plane comprises all binary zeros, and a second state indicating that a bit plane does not comprise all binary zeros, wherein the bit plane of the most significant bit (MSB) is encoded with the code word in its first state when the bit plane MSB comprises all binary zeros; (b) proceeding from the bit plane MSB to the bit plane of the least significant bit (LSB), decoding the code word in its first state for each successive bit plane comprising all the binary zeros, if any blueprint planes are present. successive bits, until reaching a first bit plane that does not include all binary zeros; and (c) decoding the code word in its second state, for the first bit plane, then using a first table for entropy coding to decode at least one code word that follows the code word in its second state; wherein at least one code word is obtained from a first table for entropy coding according to the bits in the first bit plane s.
8. The method according to claim 7, wherein: the code word having the first and second states is a single bit code word.
9. The method according to claim 7, wherein: the first table for entropy decoding does not include a multi-bit code word for decoding a bit plane with all binary zeros.
10. 10. The method according to claim 7, comprising the additional step of: providing a second table for entropy decoding to decode at least one bit plane following the first bit plane, wherein: the second table for entropy decoding includes a multi-bit code word for decode a bit plane with all binary zeros.
11. 11. The method according to claim 7, wherein: the transformation coefficient data comprises data for discrete cosine transformation (DCT).
12. The method according to claim 7, wherein: the data for the transformation coefficient comprises image data.
13. 13. A digital signal that carries data for efficient coding of a plurality of bit planes in which the data is carried for transformation coefficient, comprising: (a) a code word having a first state indicating that a bit plane comprises all binary zeros, and a second state indicating that a bit plane does not comprise all binary zeros; wherein: the bit plane of the most significant bit (MSB) is encoded with the code word in its first state when the bit plane MSB comprises all the binary zeros; and proceeding from the bit plane MSB to the bit plane of the least significant bit (LSB), each succeeding bit plane comprising all the binary zeros is encoded with the code word in its first state, if any bit planes are present successive, until a first bit plane is reached that does not include all binary zeros; (b) the code word in its second state for encoding the first bit plane; and (c) at least one code word after the code word in its second state obtained from a first table for entropy coding according to the bits in the first bit plane.
14. The signal according to claim 13, wherein: the code word having the first and second states is a single bit code word.
15. 15. An apparatus for efficient coding of a plurality of bit planes in which the coefficient data for transformation is carried, comprising: (a) means for providing a code word having a first state indicating that a plane of bits comprises all binary zeros, and a second state indicating that a bit plane does not comprise all binary zeros; (b) means for encoding the bit plane of the most significant bit (MSB) with the code word in its first state when the bit plane MSB comprises all the binary zeros; (c) means for encoding each successive bit plane comprising all binary zeros with the code word in its first state, if any successive bit planes are present, proceeding from the bit plane MSB to the bit plane of the bit less significant (LSB), until reaching a close-up that does not include all binary zeros; and (d) means for encoding the first bit plane with the code word in its second state, followed by at least one code word obtained from a first table for entropy coding according to the bits in the first bit plane.
16. 16. The apparatus according to claim 15, wherein: the code word having the first and second states is a single bit code word.
17. The method according to claim 15, wherein: the first table for entropy coding does not include a multi-bit code word for encoding a bit plane with all binary zeros.
18. The method according to claim 15, further comprising: a second table for entropy coding to encode at least one bit plane that follows the first bit plane; wherein: the second table for entropy coding includes a multi-bit code word for encoding a bit plane with all binary zeros.
19. 19. An apparatus for decoding a plurality of bit planes in which data is carried for transformation coefficient, comprising: (a) a decoding function for a code word having a first state indicating that a bit plane it comprises all binary zeros and a second state that indicates that a bit plane does not comprise all binary zeros, where the bit plane of the most significant bit (MSB) is coded with the code word in its first state when the plane of MSB bits comprises all binary zeros; (b) means for decoding the code word in its first state for each successive bit plane comprising all binary zeros, if any successive bit planes are present, proceeding from the bit plane MSB to the bit plane minus bit significant (LSB), until reaching a first bit plane that does not include all binary zeros; and (c) means for decoding the code word in its second state for the first bit plane, then using a first table for entropy decoding to decode at least one code word that follows the code word in its second state; wherein the at least one code word is obtained from a first table for entropy coding according to the bits in the first bit plane.
20. 20. The apparatus according to claim 19, wherein: the code word having the first and second states is a single bit code word.
21. The apparatus according to claim 19, wherein: the first table for entropy coding does not include a multi-bit code word for decoding a bit plane with all binary zeros.
22. 22. The apparatus according to claim 19, further comprising: a second table for entropy decoding to decode at least one bit plane that follows the first bit plane; wherein: the second table for entropy coding includes a multi-bit code word for decoding a bit plane with all binary zeros.

    SUMMARY OF THE INVENTION A system for efficient bit-plane coding of transform coefficient data is described, such as the DCT data used in a system for video coding. Decimal values for the transformation coefficients, for example, in a block of various coefficients, are converted to binary values, where each bit occupies a corresponding bit plane, from the most significant bit to the least significant bit. A bit from each coefficient is provided in a common bit plane. A single bit flag or word code (such as "0") is used to encode one or more initial all-zero bit planes, while another single bit flag (such as "1") is used to designate the first plane without any subsequent zero. For the foreground without all zero, a table for reduced coding is used to provide code words that follow the single bit flag. The table for coding is reduced in size since it does not require a special "all-zero" code word. Additionally, the use of a single bit flag to designate the initial all-zero bit planes reduces the required number of bit encoders with respect to the prior art schemes which require multi-bit all-zero code words. An encoder (200) includes a code function "242", a code function "24", a reduced table (246) and conventional tables (248). A corresponding decoder (400) includes a code word function "0" (442), a code word function "1", a reduced table (446) and conventional tables (448).