RU2406259C2 - Methods and systems for coding of clarification coefficients in video data compression - Google Patents

Abstract

FIELD: information technologies.

SUBSTANCE: method is proposed for coding of clarification coefficients in layer of scalable increase of signal-noise ratio (SNR) quality in compressed sequence of video frames. Sequence of video frames is received. Forecasting of initial video signal is composed of video sequence in current frame. Differential signal is generated by subtraction of initial video signal forecast from initial video signal in current frame. Conversion is applied to differential signal. Multiple conversion ratios are quantised. Clarification coefficient is displayed into ternary clarification symbol. Clarification symbols are grouped in a certain sequence of coding. Groups of clarification symbols are coded using codes of alternate length.

EFFECT: improved efficiency of video data coding.

30 cl, 15 dwg

Description

Related Applications

This application claims the priority of the provisional US patent application serial number 60/786800, registered March 27, 2006, entitled "Encoding refinement coefficients for fine-grained scalability," and the US provisional patent application serial number 60/830199, registered July 11, 2006, entitled "Methods and Systems for Encoding Clarification Factors for Compressing Video Data", which are incorporated herein by reference.

Technical field

The present disclosure relates generally to video processing. More specifically, the present disclosure relates to methods and systems for encoding and decoding information related to video compression methods.

State of the art

Computer technology continues to evolve in video compression techniques. Video compression refers to digital video compression. Video compression is desirable for efficiently encoding video data in video file formats and streaming and broadcasting video data formats. Compression is the conversion of data to a format that requires fewer bits, and is usually performed so that data can be stored or transmitted more efficiently. The decompression, which is the opposite of compression, produces an exact copy of the source data.

Currently, for a variety of video compression methods, a certain type of discrete transform, such as discrete cosine transform (DCT), is used to reduce the remaining spatial redundancy in the residual part of the prediction, which is performed using either temporal or spatial prediction techniques. Many video compression methods also include quantization, which can be either scalar quantization or vector quantization. After quantization, the quantized coefficients can be statistically encoded and placed in a compressed video bitstream. Then the compressed bit stream is sent to a decoder, which decompresses the bit stream and restores a good approximation of the original video data.

Video compression methods can benefit users of various forms of multimedia. For example, in the broadcast technique, digital television is in practice implemented by compressing video data. Television stations can broadcast not only for HDTV (High Definition Television), but also over multiple virtual channels on the same physical channel. Broadcasting digital video can use the MPEG-2 video compression format. H.264 / MPEG-4 and VC-1 are starting to emerge as new standards in the field of video compression.

Unfortunately, known systems and methods for encoding information in video compression suffer from various disadvantages. Accordingly, by improving systems and methods for encoding information regarding video compression, benefits can be obtained. This description describes some systems and methods for encoding information in video compression.

Brief Description of the Drawings

1 illustrates one embodiment of a scalable structure of a video encoding bit stream;

FIG. 2 illustrates one embodiment of a base layer and a plurality of quality improvement layers sent from an encoder to a decoder through a network; FIG.

figure 3 illustrates a method of encoding information;

4 illustrates one embodiment of a method for decoding an encoded sequence of video frames;

5 illustrates one embodiment of an image frame divided into a plurality of macroblocks;

6 illustrates one embodiment of the classification of significant and refinement coefficients;

7 illustrates the disadvantage of one refinement coding scheme in variable length coding (VLC);

FIG. 8 illustrates one embodiment of a subband based coding order;

FIG. 9 illustrates one embodiment of qualifier characters that may be associated with syntaxes used in qualifier encoding; FIG.

10 illustrates one embodiment of a variable length code (VLC) codebook;

11 illustrates one embodiment of a plurality of fine grain scalability (FGS) layers;

12 illustrates one embodiment of a solution to a problem caused by a grouping of refinement coefficients when the meaningful and refinement coding are interleaved;

FIG. 13 illustrates another embodiment where meaningful coding and refinement coding can be interleaved and the FGS coding order of the subbands is used;

Fig. 14 illustrates one embodiment of a decoder based decision method;

15 illustrates one embodiment of an encoder based decision method;

16 illustrates a block diagram illustrating the main hardware components commonly used for processing digital images and / or digital video data.

Detailed description

A method for encoding refinement coefficients in a layer of scalable enhancement of the signal-to-noise ratio ("SNR") quality of a compressed sequence of video frames is disclosed. Accept a sequence of video frames. From the sequence of video frames make up the prediction of the original video signal in the current frame. A difference signal is generated by subtracting the prediction of the original video signal from the original video signal in the current frame. A conversion is applied to the difference signal. Quantize the set of transform coefficients. The refinement coefficient is mapped to a ternary refinement symbol. The qualifying characters are grouped in a specific coding order. Groups of qualifying characters are encoded using variable length codes.

For different quality improvement layers, an encoding table is adaptively selected with a variable length code. In an embodiment, a coding table with a variable-length code is adaptively selected based on a probabilistic estimate of the refinement symbols. The selection of a particular coding table with a variable length code can be based on the level of the quality improvement layer. In an embodiment, the selection of a particular coding table by a variable-length code is based on whether the coefficient is refined for the first time or whether this coefficient was refined in the previous layer.

In an embodiment, the method interleaves significant coefficients and refinement coefficients. A clipping that includes many blocks is identified. Many blocks can be viewed before grouping refinement coefficients. The order in which the refinement coefficients will be encoded is determined. The refinement coefficients may be stored in the queue in accordance with the order that has been determined.

In an embodiment, each fixed number of consecutive refinement coefficients is encoded together using a variable length code table. In an embodiment, the refinement clipping coefficients are grouped together. An offset operation may be performed. The variable length code coding table includes a variable length code coding table with 27 elements, where the group length is three. In the enhancement layer, a subband based coding order is used.

An electronic device is also disclosed that implements a method for encoding refinement coefficients in a layer of scalable quality improvement of the SNR of a compressed sequence of video frames. The electronic device includes a processor and a storage device in electronic communication with the processor. Commands stored in the storage device are contained. Accept a sequence of video frames. Prediction of the original video signal in the current frame is made up of a sequence of video frames. A difference signal is generated by subtracting the prediction of the original video signal from the original video signal in the current frame. A conversion is applied to the difference signal. Quantize the set of transform coefficients. The refinement coefficient is mapped to a ternary refinement symbol. The qualifying characters are grouped in a specific coding order. Groups of qualifying characters are encoded using variable length codes.

A computer-readable medium is also disclosed, including instructions for implementing a method for encoding refinement coefficients in a layer of scalable SNR quality improvement of a compressed sequence of video frames. Accept a sequence of video frames. From the sequence of video frames make up the prediction of the original video signal in the current frame. A difference signal is generated by subtracting the prediction of the original video signal from the original video signal in the current frame. A conversion is applied to the difference signal. Quantize the set of transform coefficients. The refinement coefficient is mapped to a ternary refinement symbol. The qualifying characters are grouped in a specific coding order. Groups of qualifying characters are encoded using variable length codes.

A method for decoding refinement coefficients in an encoded sequence of video frames is also disclosed. An encoded sequence of video frames is received. The refinement symbols are decoded using a variable length coding table. The refinement coefficients are decoded using the refinement symbols, while the refinement coefficients are grouped in a specific coding order. Many transform coefficients are inverse quantized. An inverse transform is applied to the refinement signal. Compose a sequence of video frames.

In an embodiment, the encoded video frame stream includes interleaved significant symbols and refinement symbols. The location of the requested refinement coefficients is stored. The sign of the corresponding coefficient of the base layer is maintained. The counter, which is the number of requests made, is incremented. In an embodiment, it is determined when the counter reaches a predetermined count value. The read operation is started, causing normal decoding with a code table of variable length code. Compose a group of qualifying characters.

Various embodiments of the disclosure will now be described with reference to the drawings, in which like reference numerals indicate identical or functionally similar elements. Embodiments of the present disclosure, as generally described herein and illustrated in the drawings, may be embodied and constructed in a wide variety of different configurations. Thus, the following more detailed description of several embodiments of the present disclosure, as presented in the drawings, is not intended to limit the scope of the disclosure as claimed, but merely reflects representative embodiments of the disclosure.

Many features of the embodiments disclosed herein may be implemented as computer software, electronic equipment, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various components will be described generally in terms of their functionality. Whether such functionality is implemented in the form of hardware or software may depend on the particular application and design constraints imposed on the entire system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.

Where the described functionality is implemented as software, such software may include any type of machine instruction or computer-executable code stored in a memory device and / or transmitted as electronic signals via a system bus or network. Software that implements the functionality associated with the components described in this description may contain a single command or multiple commands and may be distributed through several different code segments, among various programs, and through several memory devices.

After performing general video compression methods, such as prediction, conversion, and quantization, the next step performs statistical coding on the syntax elements and quantized coefficients to form a bit stream. A reverse process is performed in the video decoder. Syntax elements and quantized coefficients are decoded from the bitstream. Then, the steps of inverse quantization and inverse transformation are performed. Finally, a reconstructed remainder is added to the prediction to reconstruct the video data.

The Joint Video Imaging Group (JVT), which consists of video encoding experts from ISO / IEC MPEG and ITU-T VCEG, is working on expanding scalable video encoding (SVC) for H.264 / AVC. Participants use standard software called JSVM (Integrated Scalable Video Data Model). JSVM supports various scalability features. The SVC bitstream can have fine-grained scalability, spatial scalability, temporal scalability, or any combination of these three types of scalability.

Fine-grained scalability, hereinafter referred to as FGS, is an important and highly desirable feature for encoding video data. This feature allows you to almost arbitrarily shorten the compressed bitstream of video frames, while at the same time providing the end user with a gradual deterioration in the quality of the decoded video signal. Since the video decoder receives and decodes a larger number of bits in this layer, this will produce better video quality. In addition, FGS in JSVM was designed in such a way that the improvement in video quality is evenly distributed throughout the video frame, as opposed to being concentrated within only a certain spatial area of the frame, since the latter case is visually undesirable for the final viewer.

FGS achieves a gradual improvement in video quality using bit-plane coding. If the video frame of the main or previous layer is quantized using the quantization parameter of the PC, then the video data of the current quality improvement layer is quantized using the quantization parameter, usually approximately PC-6, which effectively halves the quantization step. As a result of the smaller size of the quantization step in the quality improvement layer, a larger number of transform coefficients becomes nonzero. All nonzero coefficients in the quality improvement layer can be classified into three types: insignificant coefficients, significant coefficients, and refinement coefficients. The term “non-significant” coefficients refers to those coefficients that are zero in the quality improvement layer, and the corresponding coefficients in the main or previous layer are zero. The term “significant” coefficients refers to those coefficients that become non-zero in the quality improvement layer, and the corresponding coefficients in the main or previous layer are zero. The term “refinement” coefficients refers to those coefficients that already differ from zero (that is, are significant) in the main or previous layer. Their values will be specified in the current layer of quality improvement.

Since the refinement coefficients have different statistics from other coefficients, they are separately encoded in the FGS layer. Like H.264 / AVC, the SVC draft recommendation also supports two statistical coding modes: context-based adaptive binary arithmetic coding (CABAC) mode and context-based adaptive variable-length coding (CAVLC) mode. Details are provided for a scheme for improved coding of refinement coefficients in the CAVLC mode, which, as shown, not only improves coding efficiency, but also significantly reduces the complexity of the implementation of both the encoder and decoder.

In the FGS layer, refinement coefficients are sent using two syntax elements: coeff_refinement_flag and coeff_refinement_direction_flag. The first syntax element, coeff_refinement_flag, determines whether the level of the refinement coefficient is really 0. Without loss of generality, it is assumed that if coeff_refinement_flag is 0, the value of the level of the refinement coefficient is 0, while coeff_refinement_flag equal to 1 determines that the level value of the refinement coefficient is either equal to -1 or equal to 1. The second element, coeff_refinement_direction_flag, determines whether the sign of the refinement level of the conversion coefficient is the same or different from the sign of its representation of the main or previous layer. Without loss of generality, it is assumed that if coeff_refinement_direction_flag is 0, the sign of the refinement level of the transform coefficient is equal to the sign of the corresponding coefficient in the main layer; otherwise, the qualifying level has the opposite sign.

In CAVLC mode, one way of encoding the refinement coefficients in the FGS layer is to process the refinement coefficients block by block. As soon as the video encoder starts to encode the refinement coefficients in a block, all the refinement coefficients in the current block are encoded before the video encoder starts to encode other coefficients. Such a scheme was once adopted in one version of JSVM.

The table provides such a block-based refinement coding process, summarized in the following pseudo-code.

// coding of the refinement level
// vlcTab1 is a VLC table to be used to encode the refinement layer until
while (the refinement coefficient still remains in the current block) {
get the following refinement coefficient in a zigzag fashion
let m = coefficient level
encode level m using current vlcTab1
update refinement level statistics
update vlcTab1 value
}
shift bits of the qualifier level remaining in the bit buffer
// coding of the attribute of the qualifying direction
// VlcTab2 is a VLC table to be used to encode the refinement direction until
while (the refinement coefficient still remains in the current block) {
get the following refinement coefficient in a zigzag fashion
if (the current qualification level is nonzero)
{
if (the level has the same sign as the main coefficient)
let the refinement direction s = 0
otherwise
let the refinement direction s = 1
encode s direction using current vlcTab2
update refinement statistics
update vlcTab2 value
}
}
shift bits of the refinement direction remaining in the bit buffer

The refinement decoding process is performed in the reverse order in order to restore the coeff_refinement_flag and syntax coeff_refinement_direction_flag. With these two features, together with the coefficients of the main or previous layer, the decoder can reconstruct the refinement coefficients. The decoder also maintains and updates statistical information in buffers so as to remain in sync with the encoder.

In the table, vlcTab (vlcTab1 or vlcTab2) can take one of three values indicating different ways of performing VLC encoding on the syntax of qualifying features. If vlcTab = 0, refinement syntaxes are sent bit by bit (attribute coding). If vlcTab = 1, refinement syntaxes are sent in groups of three bits. If vlcTab = 2, refinement syntaxes are sent in groups of four bits. The value of vlcTab is adaptively determined in accordance with the accumulated statistical data of refinement coefficients in the clipping of the current video data. Also, the encoder and decoder monitor the value of the most probable symbol (BVS) and translate into another state the actual symbol to be encoded if the BVS changes from 0 to 1 or vice versa.

The refinement coding method of the current CAVLC may have relatively high computational complexity. For example, for a widespread method, it may be necessary to maintain and update the accumulated statistics of both qualifying syntaxes (level syntax and direction syntax). For the widespread method, it may be necessary to update the vlcTab and BVS values. For the widespread method, it may be necessary to ensure that the probability account values for 0 and 1 will not go beyond when they are scaled back when necessary. Adaptively changing group lengths can also complicate implementation.

Even with its relatively high complexity, widespread precoding can be inefficient. The simulations show that, for some sequences, CAVLC refinement coding can take up to twice the number of bits (100% degradation) compared to CABAC refinement coding, while the total degradation in the data rate of using CAVLC mode instead of CABAC is usually only about 15% or less. This inefficiency is largely due to the fact that the CAVLC refinement coding performs two cycles of bit shifting operations at the end of each block (one for the level syntax and the other for the direction syntax). If a block contains very few refinement coefficients (for example, only 1 or 2 coefficients), then frequent offset operations may make it impossible to group 3 or 4 bits together, therefore, causing large loss of bits, because in accordance with this scheme, variable-length coding is used can not.

As used herein, the terms “one embodiment”, “embodiment”, “embodiments”, “this embodiment”, “these embodiments”, “one or more embodiments”, “some embodiments”, “ certain embodiments "," one embodiment, "another embodiment" and the like mean "one or more (but not necessarily all) embodiments of the disclosure (s) disclosed" unless explicitly stated otherwise.

The term “defining” (and its grammatical variations) is used in the broadest sense. The term “defining” covers a wide variety of actions, and therefore the term “defining” can include computing, working with a computer, processing, deriving, researching, searching (for example, searching in a table, database or other data structure), establishing and etc. Also, the term “determining” may include receiving (eg, receiving information), accessing (eg, accessing data in a storage device), and the like. In addition, the term “determining” may include resolution, selection, selection, establishment, and the like.

The phrase “based on” does not mean “based only on”, unless expressly defined otherwise. In other words, the phrase “based on” describes both “based only on” and “based at least on”.

FIG. 1 illustrates one embodiment 100 of implementing a scalable video coding bitstream structure, hereinafter referred to as fine grain scalability (FGS). In the FGS, a sequence of video frames is compressed into a single stream of video frames with at least two layers: a base layer 106 and a quality enhancement layer 104. As illustrated, the FGS may further include N quality enhancement layers 102.

In one embodiment, the core layer 106 is a non-scalable encoding of a sequence of video frames and typically has the lowest quality of the sequence of video frames. The quality enhancement layer 104 and N quality enhancement layers 102 have varying quality levels that are higher than the quality level of the base layer 106. Decoding of each additional quality enhancement layer encoded after the base layer 106 results in an increasing quality of the sequence of video frames. In addition, the FGS bit stream has one significant advantage in that, during or after encoding, the bits relating to the enhancement layer 104 and N enhancement layers 102 can be shortened (reducing the bit rate and thus the quality of the decoded image ) in real time. This provides flexible control over the transmission rate of the transmitted bits and allows you to adjust the layer 104 enhance the quality and N layers 102 enhance the quality so that they match the available bit rate.

FIG. 2 is a block diagram 200 of a network illustrating a base layer 202 and N quality enhancement layers 204 sent from an encoder 208 through a network 210 to a decoder 212. Decoder 212 may decode and use the base layer 202 and all or parts of one or more layers 204 quality improvement. The decoder 212 creates an optimal approximation 214 of the original sequence 206 of video frames. In one embodiment, the optimal approximation 214 is displayed on the display 216, such as a computer monitor, television receiver, PDA (personal digital assistant), movie projector, etc. The sequence of video frames 206 is encoded as a base layer 202 and one or more quality enhancement layers 204. The lower quality base layer 202 represents a minimum quality standard, and each of the quality enhancement layers 204 is a refinement regarding the image quality of the base layer 202. In one embodiment, the video frame sequence 206 is encoded using the FGS video encoding format, as explained in FIG. 1. The FGS encoding format can be particularly useful for streaming and broadcast video applications where the available transmission bandwidth may not be known in advance. During or after coding, the quality improvement layers 204 may be arbitrarily shortened to match the available bit rate. When the available bit rate changes, the enhancement layers 204 can be shortened accordingly in real time to provide a decoded sequence of optimal quality video frames relative to the current bandwidth. For example, network 200 may shorten the quality improvement layers 204 and then send bit streams to different destinations with different bit rates.

3 illustrates a method 300 for encoding information. In one embodiment, the information to be encoded comprises a sequence of video frames. The method 300 may be performed in a video encoder, and the inversion of a method 300 may be performed using a video decoder. An encoder is a device used to change a signal or data into a code. Then the decoder can produce the original signal or approximation of the original signal from this code. The encoder and decoder may together be referred to as a video codec, which encodes the original image or sequence of video frames into a compressed form and decodes it to produce a copy or approximation of the original sequence.

The method 300 begins 302 with the receipt by the encoder 304 of a sequence of video frames. In one embodiment, the sequence of video frames is a decompressed video signal. The encoder begins to compress the sequence of video frames, using the similarities between adjacent frames of video data and making 306 prediction of the current video signal. Prediction is formed from one or more previous or future frames and is improved by compensating for differences between frames. Then, a difference signal 308 is generated by subtracting the predicted frame from the actual current frame. The encoder uses similarities between adjacent samples of the residual prediction part to reduce spatial redundancy.

The method 300 continues by applying 310 transforms to the difference signal to reduce the remaining spatial redundancy. The transformation converts the signals of the residual part of the prediction to another region in which they are represented by the conversion coefficients. The coefficients are quantized 312 to remove insignificant values, leaving a small number of significant coefficients that provide a more compact representation of the residual samples. In one embodiment, the coefficients are quantized using PC-6, which effectively halves the size of the quantization step. As a result of the reduced size of the quantization step in the current frame, more than the transform coefficients become nonzero. For example, the prediction residual encoded using method 300 may be in a quality improvement layer. If the previous layer of the video data frame or the main layer of the video data frame was quantized 312 using the PC quantization parameter, then the current quality improvement layer is quantized 312 using PC-6.

In one embodiment, nonzero quantized coefficients in the enhancement layer may be classified as insignificant, significant, or refinement coefficients. The term “non-significant” coefficients refers to those coefficients that are zero in the main or previous layer and remain zero in the current quality improvement layer. The term “significant” coefficients refers to those coefficients that are zero in the main or previous layer, but become non-zero coefficients in the current quality improvement layer. The term “refinement” coefficients refers to those coefficients that are already non-zero in the main or previous layer. In one embodiment, each refinement coefficient is mapped 313 to a ternary refinement symbol. The trinity qualifier can have one of three meanings. The encoder scans the current quality improvement layer and groups 314 refinement symbols in a specific coding order. In one embodiment, the encoder groups 314 consecutive qualifying symbols in a coding order based on subbands. Each group of qualifying characters is referred to as a "group of qualifying characters". The encoder then encodes the 316 refinement symbol groups and adds these codes to the bitstream. In one embodiment, an encoder encodes a 316 group of qualifier symbols using a variable length code (VLC) from a VLC table. In another embodiment, the VLC table used to encode one group of refinement symbols may be dynamically selected from the set of VLC tables based on the current estimate of the statistical distribution of the refinement coefficients.

In a further embodiment, the encoder may encode qualifying symbols from a plurality of enhancement layers. For example, an encoder may encode multiple enhancement layers in an FGS encoding format. In one embodiment, refinement symbol groups from a plurality of FGS enhancement layers are encoded using adaptive VLC table selection. In another embodiment, the VLC table is fixed for one FGS layer, but may be different from the VLC table used in another FGS layer.

FIG. 4 illustrates one embodiment of a method 400 for decoding an encoded sequence of video frames. The decoding method 400 may be an inverse of the encoding method 300 shown in FIG. Decoding method 400 starts 402 when the decoder receives 404 an encoded sequence of video frames. An encoded sequence of video frames may also be referred to as a bitstream. The decoder decodes 406 groups of qualifying symbols using the VLC table. If the refinement symbol groups are encoded using adaptive VLC tables based on the statistical distribution of the refinement coefficients, the same statistical information is used in the decoder to select the correct VLC table. If the decoder receives encoded refinement symbols from a plurality of enhancement layers, the decoder may adapt the VLC table used to decode the refinement symbols in accordance with their respective enhancement layers.

The method 400 continues, and the decoder decodes the refinement coefficients 408 using newly decoded refinement symbols. The decoder then proceeds to inverse quantization 410 of the decoded refinement coefficients before adding them back to the inverse quantized coefficients of the base or previous layer. Method 400 continues by applying inverse transform 412 to the inverse quantized coefficients. In one embodiment, the inverse transform may include an inverse discrete cosine transform. After the decoder applies 412 inverse transform to the refinement coefficients, the decoder creates 414 an approximation of the original sequence of video frames as it existed before the encoding method 300 of FIG. 3.

FIG. 5 illustrates one embodiment of a video frame 500 divided into multiple macroblocks. Macroblocks are the basic data elements in the decoding process. As illustrated, the video frame 500 may be divided into a plurality of macroblocks of a fixed size, each of which includes a sample of the video frame 500. In one embodiment, each macroblock includes a sample of 16 × 16 pixels (minimum image elements). In a further embodiment, each macroblock includes a luminance component (luminance) and two components of chromatic data (chrominance). The luminance component refers to the brightness of a particular sample, while the chroma components relate to the color characteristics of the sample. In one embodiment, the video frame 500 is further divided into a plurality of cuttings, cut A 502, cut B 504 and cut C 506. A cut is a sequence of macroblocks that are processed sequentially in a specific coding order. For example, clipping A 502 and macroblocks contained in clipping A 502 can be encoded and sent to a decoder before clipping B 504 and clipping C 506. In a further embodiment, the order of encoding various clippings 502, 504 and 506 can be arbitrary . For example, cut B 504 may be encoded and sent to a decoder before other cuts. Similarly, a C 506 notch can be encoded and sent to a decoder before the other two notches. Each macroblock in N quality improvement layers corresponds to a macroblock in each of the N-1 quality improvement layers and a base layer.

6 illustrates an example of how to classify coefficients. A base layer coefficient block 602 and a quality improvement layer coefficient block 604 are provided. The empty spaces in the main layer 602 and the quality improvement layer 604 represent coefficients of zero. These ratios are referred to as insignificant ratios. Each of the remaining spaces, 606, 608, 608a, and 608b, represents a coefficient that may have a nonzero value. In one embodiment, the base layer coefficient element 602 is in a previous quality improvement layer using the FGS video encoding format. As previously explained, during encoding according to the method 300 shown in FIG. 3, the transform coefficients are quantized. After quantization, nonzero coefficients in block 604 of the coefficients of the enhancement layer, for which the corresponding coefficients in block 602 of coefficients of the base layer are zero, are referred to as significant coefficients 606. Those coefficients for which the corresponding coefficients in the main layer are nonzero are hereinafter referred to as refinement coefficients 608. Since the significant coefficients 606 and the refinement coefficients 608 have different statistical data, they are encoded separately in the quality improvement layer 604, researching the FGS coding format.

In one embodiment, the significant coefficients 606 are those coefficients for which the corresponding coefficients 606a and 606b (a coefficient representing the same subband) in the main or previous layer are zero. The refinement coefficients 608 are those coefficients for which the corresponding coefficients 608a and 608b in the main or previous layer are nonzero.

7 illustrates a drawback 700 of the widespread CAVLC refinement coding scheme that uses block coding order. Block order essentially generates a spatially uneven quality improvement within the frame 710 of the video data. As shown in FIG. 7, when the 708 bit stream is shortened, all video blocks presented in the 708 bit stream before shortening 704 (blocks 0, 1, ... n-1) will have better quality 712 video data compared to blocks 706 video data (blocks n and on) in which the refinement coefficients are lost during shortening 714. The best order for encoding these refinement coefficients is to use a subband-based coding order, which is shown in detail in the following drawing.

FIG. 8 illustrates one embodiment of subband-based order 800 coding. In one embodiment, the first block 802, the second block 804, and the third block 806 are three coefficient blocks in the same quality improvement layer. In one embodiment, blocks 802, 804, and 806 are blocks in a first FGS enhancement layer, so that subband-based FGS encoding order 800 can be extended to and used in the first FGS layer. Compared to the block-based order, the subband-based coding order 800 sends refinement coefficients in accordance with the sub-band of the transform domain to which they belong. In one embodiment, the Bx, y notation illustrated in blocks 802, 804, and 806 may represent a coefficient in the subband y and in block x. For example, B0.0 represents a coefficient in a first subband (subband 0) and in a first block 802.

In one embodiment, the encoder first encodes any refinement coefficients present at positions B0.0, B1.0, and B2.0. The encoder then encodes any refinement coefficients present at positions B0.1, B1.1 and B2.1, etc. The subband coding order 800 facilitates the coding and sending of refinement coefficients according to the subband of the transform domain to which they belong. Using the order of 800 coding of the subbands results in a more uniform spatial improvement in quality, since more refinement bits are received.

FIG. 9 illustrates one embodiment 900 of refinement symbols (ref_symbol) 906 that may be associated with two syntaxes used in the refinement coding, coeff_refinement_flag 902 and coeff_refinement_dir_flag 904. As previously explained, if the refinement coefficient is 0, the syntax element coeff_refinementfl a value of 0. Accordingly, ref_symbol 906 is also set to 0. If the refinement coefficient has a non-zero value, coeff_ref_flag 902 is set to 1. If the refinement coefficient has the same sign as the corresponding the first refinement coefficient in the main layer or the previous enhancement layer, coeff_ref_dir_flag 904 is set to 0. If the sign of the refinement coefficient in the main layer or the previous enhancement layer is opposite to the sign of the refinement coefficient in the current layer to be encoded, coeff_ref_dir_flag 904 is set to 1. In another embodiment the implementation of the parameters that determine the value of coeff_ref_dir_flag 904 are opposite.

If coeff_ref_dir_flag 904 is set to 0, ref_symbol 906 is set to 1, and if coeff_ref_dir_flag 904 is set to 1, ref_symbol 906 is set to 2. Then the values of ref_symbol 906 are encoded and included in the bit stream. One embodiment of encoding ref_symbol 906 values is provided below.

FIG. 10 illustrates one embodiment of a variable length coding table (VLC) 1000 used to encode reference symbol values, as explained in FIG. 9. In the coding order of the subbands, consecutive qualifier symbols may be grouped together and the qualifier symbol group 1002 is encoded using the VLC table 1000. Table 1000 includes many elements. Each element has the value of a qualifier group 1002, as explained in FIG. 9, a code length 1004 for a variable length code for encoding a particular character group 1002 and a variable code codeword 1006 for encoding a specific character group 1002. Then, the codeword 1006 is included in the bitstream and sent to the decoder.

In one embodiment, a refinement coding uses a single VLC table similar to table 1000; it is not necessary for the encoder and decoder to constantly accumulate and update statistics of the refinement symbol, thus the computational complexity can be significantly reduced. In another embodiment, the VLC table is selected from a plurality of tables based on an estimate of the statistical distribution of refinement coefficients, efficiency can be improved with the added complexity of probabilistic estimation and table selection.

Turning now to FIG. 11, it is noted that current SVC standards allow for multiple layers on top of a discrete base layer 1102. When there are more than one FGS layer 1104 and 1106, starting from the second FGS layer 1106 and above, all refinement coefficients belong one of two classes: “old” refinement coefficients are coefficients that already have refinement coefficients in the previous layer 1104 FGS, and “new” refinement coefficients are coefficients that became significant only in the previous layer The second FGS 1104 and only enter the refinement stage in the current layer 1106. FIG. 11 provides an example of the presence of two FGS layers 1104 and 1106. The first FGS layer 1104 contains significant coefficients 1108 as well as the refinement coefficients 1110. The second FGS layer 1106 includes the "old" refinement coefficients 1112, as well as the "new" refinement coefficients 1114. The coefficient 1116 represents a new significant coefficient in the second layer 1106 FGS. The remaining unfilled positions are insignificant coefficients that do not have non-zero values.

Two types of refinement coefficients 1112 and 1114 have different probability distributions of symbols. Therefore, in one embodiment, it is better to use different VLC tables to encode them. In addition, depending on the level of the current FGS layer, different VLC coding tables may be used. For example, to code the refinement coefficients included in the second FGS layer 1106, a different VLC table may be used than the table used to apply the refinement coefficients included in the first FGS layer 1104. Therefore, in order to improve efficiency, a coding table better suited for such a probability distribution should be used. This adaptation of the VLC table has very low complexity because it is based solely on the class of refinement coefficients and the level of the FGS layer, and not on relying on the collection and updating of symbol statistics. In another embodiment, the VLC table may be dynamically selected based on an estimate of the statistical distribution of refinement coefficients for the current FGS layer.

12 illustrates one embodiment 1200 of solving a problem caused by a grouping of refinement symbols when meaningful and refinement coding is interleaved. In FGS coding, the significant coding pass and the refinement coding pass for different blocks are usually interleaved. One reason for this is that one block may contain fewer significant coefficients than another block. When the significant coding pass of a certain block is completed, after that, the qualifying pass for this block begins, while the block next to it still passes its significant coding pass. In one embodiment, meaningful coding (S0.0 and then CBL (end of block)) for block A 1202 ends after two cycles. Then, refinement coding R0.0 (and later R0.1) will begin. But at this time, significant coding of block B 1204 (coefficients S1,2 and KBL) is still ongoing.

Another reason for interleaved significant and refinement coding is to use the coding order of the subbands. 13 illustrates another embodiment 1300 where meaningful coding and refinement coding may be interleaved. In one embodiment, coefficients are scanned in subbands. For example, when viewing the 0th subband 1306, block A 1302 contains a significant coefficient S0.0. Therefore, significant coding of S0.0 is performed. Similarly, when the 0th subband 1306 of block B 1304 is scanned, refinement coding of the refinement coefficient R1.0 is performed.

This behavior of the interleaved significant and refinement coding produces a problem when the refinement coding is performed by grouping a plurality of (e.g., three) refinement coefficients. On the encoder side, three qualifier symbols are buffered before the VLC codeword is sent. But on the decoder side, whenever faced with decoding of the refinement coefficient, the decoder will try to read the VLC code word from the bitstream, which represents a group of three refinement symbols. But this codeword is still buffered in the encoder encoding buffer and has not yet been sent. Loss of synchronization between the encoder and the decoder means that such a coding scheme (grouping of qualifying symbols when the significant and qualifying passes are interspersed) will lead to a failure. To solve this problem, two solutions are provided and presented below.

FIG. 14 illustrates a decoder based decision method 1400 that does not require any change on the encoder side. The encoder can always try to group together every three qualifying characters before sending the keyword. On the decoding side, when a refinement coefficient is required, the following operations are performed. Method 1400 begins by setting the counter 1402 to its initial state of zero. 1404 the location of the requested refinement coefficient is stored, 1406 the sign of the corresponding coefficient of the base layer is stored, the counter (initialized to zero at the beginning of clipping decoding) is increased to track how many such requests were made 1408, if this counter reaches three 1410, the “read” operation is activated 1412 and normal VLC decoding is performed. A group of three qualifying symbols is returned, which are then distributed back to the corresponding locations of the coefficients, as they were previously stored, 1414, and the counter returns to its original position at zero 1416. If the counter has not yet reached three 1410, method 1400 returns to step 1402 and continues performance.

While the decision on the decoder side places a heavy burden on the decoder, the decision on the encoder side has minimal complexity impact on the decoder. The basic concept is that the encoder always sends a plurality (for example, three) of refinement coefficients, the decoder is guaranteed to have enough refinement characters in the buffer when refinement coefficients are required. FIG. 15 illustrates a method 1500 of one type of encoder based decision. The encoder previews 1502 a certain number of blocks in the slice and determines the order 1504 with which all refinement coefficients will be sent during the actual encoding stage. In this method, the refinement coefficients are stored 1506 in the queue in accordance with this encoding order. At the coding stage, when it is necessary to send a refinement coefficient, it is determined 1508 whether it has been grouped together with the previous coefficient (s) and has already been sent. If so, then no action is taken 1510; otherwise, this refinement coefficient is sent 1512 along with two coefficients located immediately next to it in the queue. Method 1500 continues until the encoding pass is completed.

Note that decisions based on both the encoder and the decoder are applied to any type of interleaving of a significant pass and a refinement pass from various blocks, including, but not limited to, the two coding orders currently supported in the SVC draft recommendation. These two solutions also apply to grouping any number of qualifying characters, and are not limited to only grouping three characters.

It should also be noted that both encoder-based and decoder-based solutions can be applied to block-based refinement coding as defined in the 5th Joint Project (JD5) SVC, thereby making it unnecessary to offset the remaining refinement bits stored in buffer at the end of each block. For a decoder based solution, an additional change on the encoder side can guarantee the synchronization of the encoder and decoder. Since the block-based refinement coding scheme in JD5 adapts the VLC table, the number of grouped update symbols can vary between 1, 3, and 4. At the end of the block, the encoder must pause the adaptation of the VLC table until it encounters a sufficient number of refinement factors (this will happen a little later, since the encoder performs refinement coding of consecutive blocks). When it encounters enough refinement coefficients to trigger a VLC codeword write operation, the encoder can resume normal adaptation of the VLC table. According to this scheme, the frequent offset operation (performed once for each block), as used in the widespread refinement coding, can be minimized, and coding efficiency can be significantly improved.

At the end of the encoding of a certain number of refinement coefficients, an offset operation may be used if the number of remaining refinement coefficients is nonzero and less than the size of the group used in the normal encoding process. One way to perform the offset operation is to add the most likely characters to make a complete group and encode the last group in the same way as other groups. For example, suppose that the normal group size is taken to be 3, and that there is one remaining coefficient to be encoded in the offset operation, and that the symbol is 2. The group (2, 0, 0) will be encoded, assuming that the symbol “0” is the most likely character in the ternary character set. On the decoder side, these padding symbols are discarded. Offsets can also be performed more frequently to reduce implementation complexity. A group of refinement coefficients that are encoded together is formed from various locations. Limiting the distance between the coefficients in the same group will result in several ungrouped characters that must be processed using shift operations. In another embodiment, instead of forming a complete group by adding the most probable characters that will be discarded by the decoder, VLC tables designed to encode smaller groups are used. This can be used when the number of refinement factors to be shifted is known to the decoder. Suppose the group size is 3 and there is only one “R” coefficient to be shifted. Instead of encoding a group (R, 0, 0) using VLC tables, it is more efficient to use a special VLC table to encode a group of 3 characters. For example, a qualifier with a value of "0" is encoded as "1", a qualifier with a value of "1" is encoded as "01", and a qualifier with a value of "2" is encoded as "00". Similarly, if the group size is 3, and there are two coefficients, “R1” and “R2”, to be shifted. Instead of encoding a group (R1, R2, 0), other VLC tables can be used to encode a smaller group (R1, R2). In one embodiment, Exponential-Golomb code is used to encode the group index, which is computed as R1 * 3 + R2. It should be noted that in an equivalent way, the index of the encoded group can also be formed as R1 + R2 * 3.

FIG. 16 illustrates an apparatus 1600 configured to process digital images and / or digital video data. The device 1600 may be or may be implemented in a digital television receiver, digital direct broadcast system, wireless communication device, personal digital assistant (PDA), compact portable computer, desktop computer, digital camera, digital recording device, digital television receiver with support for work in a network, radiotelephone of cellular or satellite communication, or in any telecommunication device with video telephony (PTS) capabilities.

The device 1600 may process, encode, decode, transmit, and / or receive image data and / or video data. Video data may be captured by a video camera, such as a video capture module (or image sensor) 1612, extracted from a video archive, or otherwise obtained. The video encoding module 1610 in the device 1600 may use a video encoding standard, such as MPEG-4, ITU-T H.263, ITU-T H.264, or any other video encoding standard. Video encoding module 1610 may support inter-frame coding techniques, such as motion estimation and motion compensation, and intraframe coding techniques, such as spatial estimation and intra-frame prediction coding methods.

The device 1600 may include an image / video capture device 1612, such as a camera or video camera, to capture images or sequences of video frames and store the captured images or sequences in memory 1614. The image / video processing module 1602 may process images and / or sequences of video frames . A storage device 1604 may store images and / or sequences of video frames before and after such processing.

The transceiver 1616 may receive and / or transmit encoded sequences of video frames to another device. The transceiver 1616 may use a wireless standard, such as code division multiple access (CDMA). Examples of CDMA standards include CDMA 1xEV-DO, WCDMA (Wideband Code Division Multiple Access), etc.

One or more elements of device 1600 may be communicatively coupled via communication bus 1618. In addition to, or instead of, the elements shown in FIG. 16, other elements may be included in device 1600. The architecture illustrated in FIG. 16 is merely an example. The methods described herein can be implemented using a number of other architectures.

The storage device 1614 may have a relatively large memory space. Storage device 1614 may comprise dynamic random access memory (DOS) or flash memory. The storage device 1614 may comprise a NOR or NAND logic element memory technology, or any other data storage technology. In other examples, the storage device 1614 may comprise a non-volatile storage device or any other type of storage device.

The image / video processing module 1602 may comprise a chipset for a mobile radiotelephone, which may include hardware, software, firmware, and / or one or more microprocessors, digital signal processors (DSPs), application oriented integrated circuits (ISPS) , user-programmable gate arrays (PPVM) or their various combinations. Processing module 1602 may include a local storage device 1604 connected to image / video pre-processing module 1608 and to image / video data encoding module 1610. Encoding module 1610 may comprise an encoder / decoder (codec) for encoding (or compressing) and decoding (or decompressing) digital video data.

The local storage device 1604 may comprise a less and faster memory space relative to the storage device 1614. For example, the local storage device 1604 may include synchronous dynamic RAM (DLC). The local storage device 1604 may comprise an on-chip memory integrated with other components of the processing module 1602 to provide quick access to data during an encoding process requiring intensive processor operation. However, storage devices 1614 and 1604 can be combined into one storage device or can be implemented in a number of other configurations. A storage controller 1606 can control the selection and writeback of a storage device to local storage 1604.

The image / video pre-processing module 1608 may perform one or more image processing techniques on frames of a sequence of video frames to improve image quality and thus improve the quality of the sequence of video frames. For example, image / video pre-processing module 1608 may perform techniques such as correction of a decline in lens performance, scaling, color correction, color conversion, and spatial filtering. The image / video pre-processing module 1608 may also perform other methods. In general, the methods performed by module 1608 are hereinafter referred to as “image pre-processing” methods, since these methods precede the encoding performed by the image / video encoding module 1610.

The image / video capture module 1612 may include image sensors that include color filter groups (GSCFs) made on the surface of the sensors. Image pre-processing by module 1608 can improve the quality of the sequence of video frames captured by capture module 1612. For example, preprocessing module 1608 and / or encoding module 1610 may comprise a DSP programmed to process images captured by capture module 1612. The same area of the storage device 1604 (or storage device 1614) can be used for the purposes of image pre-processing, and for other storage purposes.

The image / video encoding module 1610 may perform encoding of images and / or video data, which may include one or more video compression methods, such as inter-frame compression and / or intra-frame compression. For example, coding module 1610 may implement motion estimation and motion compensation methods to use temporal or inter-frame data correlation to provide inter-frame compression. Alternatively or in addition, coding module 1610 may perform spatial estimation and intra-frame prediction techniques to use spatial or intra-frame data correlation to provide intra-frame compression. The results of motion compensation (or intra-frame prediction) are referred to as “remainder” and may contain a data block indicative of the differences between the current video data block to be encoded and the predicted block identified by motion estimation or spatial estimation.

After the coding module 1610 performs motion compensation (or intra-frame prediction) to form a remainder, a series of additional steps can be performed to further encode the remainder and further compress the data. Additional steps may depend on the coding standard used, but are generally referred to as “residual coding”. Encoding module 1610 may perform one or more of these video compression methods to reduce the amount of data needed to communicate a sequence of video frames through transceiver 1616 to another device.

Information and signals may be represented using any of a variety of different technologies and methods. For example, data, instructions, commands, information, signals, bits, symbols and code elements that may be referred to throughout the above description may be represented by voltages, currents, electromagnetic waves, magnetic fields or particles, optical fields or particles, or any their combination.

The various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, software, or combinations thereof. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally based on their functionality. Whether functionality such as hardware or software is implemented depends on the particular application and design constraints imposed on the entire system. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present disclosure.

The various illustrative logical blocks, modules, and circuits described in connection with the embodiments disclosed herein may be implemented or performed using a general purpose processor, digital signal processor (DSP), application-oriented integrated circuit (ISPS), user programmable gate array (PPVM) or other programmable logic device, discrete logic element or transistor logic circuits, discrete hardware components, or any of their components bination, designed to perform the functions described in this description. A general purpose processor may be a microprocessor, but in the alternative, the processor may be any conventional processor, controller, microcontroller, or state machine. A processor may also be implemented as computing devices, for example, as a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors together with a DSP core, or any other such configuration.

The steps of a method or algorithm described in connection with the embodiments disclosed herein may be embodied directly in hardware, in a software module executed by a processor, or in a combination of the two. The software module may reside in RAM memory, flash memory, ROM memory, EEPROM memory, EEPROM (electrically erasable EEPROM) memory, registers, a hard disk, a removable disk, a CD-ROM, or any other form of recording medium known in the art. The recording medium is connected to the processor so that the processor can read and write information from the recording medium. Alternatively, the recording medium may be combined with a processor. The processor and the recording medium may reside in the ISPO. ISPI can reside in a user terminal. Alternatively, the processor and the recording medium may reside as discrete components in a user terminal.

The methods disclosed herein comprise one or more steps or actions for achieving the described method. The steps and / or actions of the method may be interchangeable with each other without departing from the scope of the present disclosure. In other words, if a specific order of steps or actions is not required for the functioning of the embodiment to function properly, the order and / or use of certain steps and / or actions can be changed without departing from the scope of the present disclosure.

Although certain embodiments and applications of the present disclosure have been illustrated and described, it should be understood that the disclosure is not limited to the exact configuration and components disclosed herein. Various modifications, alterations and variations that will be apparent to those skilled in the art can be made in the placement, operation and details of the methods and systems of the present disclosure disclosed herein without departing from the scope and scope of the disclosure.

Claims (30)

1. A method of encoding refinement coefficients in a layer of scalable increase in signal-to-noise ratio (SNR) of a compressed sequence of video frames, comprising
receiving a sequence of video frames,
predicting the original video signal in the current frame from a sequence of video frames,
generating a differential signal by subtracting the prediction of the original video signal from the original video signal in the current frame,
applying a transform to a difference signal, which draws the difference signal to a region in which the difference signal is represented by a plurality of transform coefficients,
quantizing a plurality of transform coefficients to form a plurality of non-zero transform coefficients, each of the non-zero transform coefficients having a non-zero value, and at least some of the non-zero transform coefficients are refinement coefficients, each of the refinement coefficients having a non-zero value before quantization,
mapping each refinement coefficient into a ternary refinement symbol to form a plurality of refinement symbols, moreover, the ternary refinement symbol has three values, and each refinement symbol has one of three values,
grouping the qualifying characters in a specific coding order to form a plurality of groups of qualifying characters, and
coding of qualifying character groups using variable length codes. 2. The method according to claim 1, further comprising using a coding table with a variable-length code adaptively selected for various quality improvement layers. 3. The method according to claim 1, further comprising using a coding table with a variable-length code adaptively selected based on a probabilistic assessment of the refinement characters. 4. The method according to claim 2, in which the selection of a particular coding table with a variable length code is based on the level of the quality improvement layer. 5. The method according to claim 4, in which the selection of a specific coding table with a variable-length code is based on whether the coefficient is refined for the first time or the coefficient was specified in the previous layer. 6. The method according to claim 1, in which significant coefficients and refinement coefficients are interleaved, the method further comprising identifying a video data clipping that includes a plurality of blocks, viewing a plurality of blocks before grouping the refinement coefficients, determining the order in which the refinement coefficients will be encoded , and storing the refinement coefficients in the queue in accordance with the order that was determined. 7. The method according to claim 1, further comprising identifying a clipping of video data that includes a plurality of blocks. 8. The method according to claim 7, further comprising scanning a plurality of blocks before grouping the refinement coefficients. 9. The method according to claim 1, further comprising determining the order in which the refinement coefficients will be encoded. 10. The method according to claim 9, further comprising storing the refinement coefficients in the queue in accordance with the order that has been determined. 11. The method according to claim 1, in which each fixed number of consecutive refinement coefficients is encoded together using a coding table with a variable length code. 12. The method according to claim 7, further containing a grouping of refinement coefficients for cutting video data together. 13. The method according to claim 1, further comprising performing a bit shifting operation. 14. The method according to item 13, in which the operation of the bit offset includes adding one or more most probable characters to one or more qualifying characters to make a complete group. 15. The method of claim 13, wherein the bit shifting operation includes using a coding table with a variable length code that is selected based on the number of refinement coefficients to be processed in the bit shifting operation. 16. The method according to claim 2, in which the coding table with a variable-length code contains a coding table with a variable-length code with 27 elements, the length of the group being three. 17. The method according to claim 1, further comprising using subband-based coding order in the quality improvement layer. 18. An electronic device for encoding refinement coefficients in a layer of scalable increase in signal-to-noise ratio (SNR) of a compressed sequence of video frames, the electronic device comprising
CPU,
a storage device in electronic communication with the processor,
instructions stored in a storage device, the instructions being executed by the processor to receive a sequence of video frames,
making forecasts of the original video signal in the current frame from a sequence of video frames,
generating a differential signal by subtracting the prediction of the original video signal from the original video signal in the current frame,
applying the transform to the difference signal, which draws the difference signal into a region in which the difference signal is represented by a plurality of transform coefficients,
quantizing a plurality of transform coefficients to form a plurality of non-zero transform coefficients, each of the non-zero transform coefficients having a non-zero value, and at least some of the non-zero transform coefficients are refinement coefficients, each of the refinement coefficients having a non-zero value before quantization,
mapping each refinement coefficient into a ternary refinement symbol to form a plurality of refinement symbols, moreover, the ternary refinement symbol has three values, and each refinement symbol has one of three values,
grouping the qualifying characters in a specific coding order to form a plurality of groups of qualifying characters, and
encoding groups of qualifying characters using variable length codes. 19. The electronic device according to claim 18, wherein the instructions are further executed to use a coding table with a variable length code adaptively selected for various quality improvement layers. 20. The electronic device according to p. 18, in which the electronic device is configured to interleave significant coefficients and refinement coefficients, and in which the commands are additionally executed to identify the clipping of video data, which includes many blocks, view multiple blocks before grouping the refinement coefficients, determining the order in which the refinement coefficients will be encoded, and storing the refinement coefficients in the queue in accordance with the order that was determined en. 21. A computer-readable medium containing processor-executable instructions for encoding refinement coefficients in a scalable signal-to-noise ratio (SNR) layer of a compressed sequence of video frames, the commands comprising receiving a sequence of video frames,
predicting the original video signal in the current frame from a sequence of video frames,
generating a differential signal by subtracting the prediction of the original video signal from the original video signal in the current frame,
applying a transform to a difference signal, which draws the difference signal to a region in which the difference signal is represented by a plurality of transform coefficients,
quantizing a plurality of transform coefficients to form a plurality of non-zero transform coefficients, each of the non-zero transform coefficients having a non-zero value, and at least some of the non-zero transform coefficients are refinement coefficients, each of the refinement coefficients having a non-zero value before quantization,
mapping each refinement coefficient into a ternary refinement symbol to form a plurality of refinement symbols, moreover, the ternary refinement symbol has three values, and each refinement symbol has one of three values,
grouping the qualifying characters in a specific coding order to form a plurality of groups of qualifying characters, and
coding of qualifying character groups using variable length codes. 22. A method for decoding refinement coefficients in an encoded sequence of video frames, comprising
receiving an encoded sequence of video frames, decoding the refinement symbols of the sequence of video frames using a coding table with a variable length code, the refinement symbols being encoded using a plurality of refinement coefficients, each refinement coefficient being mapped to a ternary refinement symbol to form refinement symbols, and the ternary refinement symbol has three values, and each qualifying character has one of three values,
decoding the refinement coefficients using the refinement symbols, the refinement coefficients being grouped in a specific coding order, and the refinement coefficients being contained in a plurality of transform coefficients,
dequantizing a plurality of transform coefficients to form a refinement signal,
applying the inverse transform to the refinement signal, and compiling a sequence of video frames using the inverse transform of the refinement signal. 23. The method according to item 22, additionally containing the use of various coding tables with variable-length code, adaptively selected for various layers of improving quality. 24. The method according to item 22, in which the encoded sequence of video frames includes alternating significant characters and refinement characters, further comprising storing the location of the requested refinement coefficient, preserving the sign of the corresponding coefficient of the base layer, and increasing the counter, which refers to the number of requests made. 25. The method according to paragraph 24, further comprising determining when the counter reaches a predetermined count, and activating a read operation and invoking the decoding of the normal coding table with a variable-length code to obtain a group of qualifying characters. 26. The method of claim 25, further comprising storing a group of qualifying symbols at appropriate locations as they were previously stored. 27. The method according to item 22, further comprising performing a bit offset operation. 28. The method of claim 27, wherein the bit shifting operation includes discarding padding symbols. 29. The method of claim 27, wherein the bit shifting operation includes using a coding table with a variable length code that is selected based on the number of refinement coefficients to be processed in the bit shifting operation. 30. An electronic device for encoding refinement coefficients in a layer of scalable increase in signal-to-noise ratio (SNR) of a compressed sequence of video frames, the electronic device comprising
means for receiving a sequence of video frames,
means for predicting the original video signal in the current frame from a sequence of video frames,
means for generating a difference signal by subtracting the prediction of the original video signal from the original video signal in the current frame,
means for applying the conversion to the difference signal, which draws the difference signal into a region in which the difference signal is represented by a plurality of conversion coefficients,
means for quantizing a plurality of transform coefficients, for generating a plurality of non-zero transform coefficients, each of the non-zero transform coefficients having a non-zero value, and at least some of the non-zero transform coefficients are refinement coefficients, each of the refinement coefficients having a non-zero value before quantization,
means for mapping each refinement coefficient into a ternary refinement symbol for generating a plurality of refinement symbols, wherein the ternary refinement symbol has three values, and each refinement symbol has one of three values,
means for grouping the qualifying characters in a specific coding order to form a plurality of groups of qualifying characters, and
means for encoding groups of qualifying characters using variable length codes.

Images