JP4575344B2 - Video coding system with multiple independent coding chains for dynamic and selective playback in reduced or full size - Google Patents

Description

  Embodiments of the present invention relate to a video encoding system that supports video playback at a variably selected size. Embodiments of the present invention relate to video coding systems that support intra-frame coding schemes, particularly as found in video editing applications.

  Modern video editing applications are software-based tools that allow an operator (editor) to generate an output video sequence from multiple input video sequence candidates. Using video editing applications, it is possible to create movies and TV shows from a variety of image sources. For example, to make an interesting conversation between two people, an editor can generate an output sequence that traverses two or more different camera angles. The editor can also select the first video sequence in one section and crossfade the second video sequence in another section. In order to support such editing functions, video editors include a graphical user interface that displays various input sequence candidates along a common time axis. The editor can also specify cuts between input streams that occur in the output stream, and can also specify any video effects (such as crossfading) that occur between input streams. These draft editing settings can be freely modified until the final settings are obtained.

  In video editing applications, you can always test your editing settings. An application typically accesses encoded video data indicating input stream candidates. The application's graphical user interface also includes a window for playback of editing settings on a portion of the display. In playback, the application decodes a plurality of stored video data sequences and applies video effects to display the video sequence. In order to draw a displayable video sequence in real time, the decoding process and the video effect process must be performed within the time of one frame (usually 1 / 30th of a second). In order to perform the decoding process as fast as possible, a conventional video editing application uses encoded video data that has been subjected only to intra prediction (intraframe prediction). That is, each frame is decoded without performing temporal prediction (prediction in the time axis direction).

  Video editing application designers may be required to handle a wider variety of encoded video data, including data encoded by temporal prediction.

  The inventors have determined that it is effective to decode the input video data from the original encoding format and re-encode the data based on the intra prediction technique in order to perform the decoding process as fast as possible. In addition, decoding and re-encoding reduce the compression rate and increase the file size compared to the input data, but enable faster decoding during the test operation.

  The inventors have also recognized that further optimization can be applied to the decoding process executed in the video setting test mode by decoding and re-encoding the input data.

  Embodiments of the present invention provide an encoding system that encodes data based on a pair of encoding chains. The first encoding chain is a self-decoded encoded video data that represents a reduced size source video sequence that is large enough to support real-time playback and display capabilities in a video editing application. Is generated. The second encoding chain generates encoded video data indicating supplemental (remaining) data. The supplemental data is decoded together with the encoded video data of the first encoding chain to obtain a full size display source video sequence. The output of the first encoding chain is stored in a file-structured memory and can be accessed independently of the second chain output, thus enabling decoding and playback in real time.

FIG. 1 is a block diagram of a video encoding system 100 in an embodiment of the present invention. Video encoding system 100 accepts video sequence data from some source, such as a video capture device or storage device. In general, image data of a video sequence is composed of a plurality of frames, and each frame includes a pixel array. Pixel data can be divided into a luminance signal and a pair of color signals (Y, C _r , C _b ), each of which can be processed independently in the system 100.

  The video encoding system 100 includes a transformer 110, a quantizer 120, a splitter 130, and a pair of processing chains 140, 150. The processing chain includes run length encoders 142 and 152 and variable length encoders 144 and 156, respectively.

  The conversion circuit 110 converts the pixel data block from the source frame into a coefficient data block according to a predetermined conversion method. For example, the converter 110 operates according to a discrete cosine transform (DCT). DCT coefficients are usually written as a two-dimensional array of coefficients. In the most general configuration, an 8 × 8 pixel block of source data is converted into an 8 × 8 array of DCT coefficients as in the array shown in FIG. Here, the coefficient at the origin (position (0, 0)) of the array represents a direct current (DC) component of the block, and the other coefficients represent various frequency components included in the block. Each coefficient along the main axis of the block indicates an image element in which the frequency in a single direction gradually increases. Coefficients along the main diagonal of the block indicate image elements whose frequencies in multiple directions are gradually increased.

The quantizer 120 truncates the coefficient by dividing by the quantization parameter (q _p ). This reduces the size of the coefficients used in the subsequent encoding process. Due to quantization, some low level coefficients are truncated to zero. The quantization parameter may be changed for each block in the frame or for each frame. In this case, information about the quantization parameter itself is included in the encoded data output from the video encoding system 100, and the quantization parameter is reconfigured at the time of decoding, and the inverse quantization process is performed.

  The splitter 130 indicates the demultiplexing process performed in the video encoding system 100 and distributes the coefficient of each block to one of the pair of encoding chains 140 and 150. In the present embodiment, for each block of coefficient data input to the splitter 130, the splitter 130 transfers a low-frequency coefficient set to the first encoding chain 140 and the remaining coefficients to the second encoding chain 150. For example, to restore a video signal that can be displayed in a quarter size of the source image, a quarter low frequency coefficient is encoded in the first encoding chain (for an 8 × 8 block). It is sufficient to encode 16 of the generated 64 coefficients) and encode the remaining coefficients in the second encoding chain. Similarly, other display sizes can be handled.

  Encoding chains 140 and 150 include run-length encoders 142 and 152 and variable length encoders 144 and 154, respectively. In one embodiment, each run-length encoder 142, 152 performs run-length encoding according to a conventional zigzag scan direction. The run-length encoder scans across the various coefficient positions starting from the origin and counts the number of positions traversed to reach a non-zero coefficient. The number of traversed positions is output as a run length, and a non-zero coefficient is output as a level. Until the entire block is scanned across, the encoder continues traversing (scanning) within the block and outputs runs and levels in succession.

  Of course, in the first chain 140, the run-length encoder 142 only needs to traverse (scan) the input coefficients. In the above example, as shown in FIG. 2B, a quarter size is displayed with a plurality of 4 × 4 blocks. A conventional zigzag scan is appropriate for such a block.

  In the second chain 150, the run-length encoder 152 scans the remaining coefficients in the block. However, in order to target a portion excluding the coefficients passed to the first chain 140 (passed to the first chain 140). In general, the block has an irregular arrangement of coefficient values (FIG. 2 (c)). In one embodiment, it is particularly effective to insert a dummy zero value at a position that would otherwise be occupied by a coefficient in a small block and then treat the remaining coefficients as a normal array. That is, the run-length encoder 152 can perform a conventional zigzag scan over the entire block. For relatively smooth (smooth textured) image content, the coefficient on the main diagonal far from the origin often has a zero value, so run-length encoding can be performed particularly efficiently. .

  Alternatively, the run length encoder 152 can scan the remaining blocks in accordance with a conventional zigzag scan in a manner that skips the positions of the coefficients passed to the first chain 140. For example, referring again to FIG. 2 (c), the run-length encoder 152 can start scanning from position (4, 0) rather than starting from position (0, 0). Since the coefficient positions (3, 1), (2, 2), and (1, 3) do not exist in the remaining blocks, the next “next” position in the zigzag scan direction is (0, 4). The next zigzag scan from position (0,5) to (5,0) is because the coefficient positions (3,2) and (2,3) are small block members (present in the small block) Similarly, these are skipped.

  In a further embodiment, the remaining blocks may be parsed into sub-blocks for independent run length encoding according to the zigzag scan direction. For example, in the embodiment of FIG. 2C, the remaining blocks can be divided into 4 × 4 arrays of three coefficient data, and run-length encoding can be performed independently for each.

  In other embodiments of the present invention, the variable length encoder may perform encoding according to one of a plurality of lossless encoding algorithms. In a lossless encoding algorithm, VLC data can be directly decoded into run-length data by directly calculating without referring to any lookup table. In particular, when the decoder is implemented by software, it is possible to perform real-time decoding faster than using a lookup table by directly calculating the source data. By way of example, the variable length encoder may operate according to a coding algorithm such as a Golomb-Rice code or an exponent Golomb code.

  The variable length encoders 144 and 154 of the two chains 140 and 150 encode the run and level pair data output from the run length encoders 142 and 152. In other embodiments, the run length encoders 142, 152 of the processing chains 140, 150 can be omitted. In this embodiment, the variable length encoders 144 and 154 directly encode the coefficient data output from the splitter 130. For this reason, the run-length encoders 142 and 152 in FIG. 1 are indicated by dotted lines.

  The encoded video data for each frame is stored in a file 160 in the storage device. In this embodiment, in order to facilitate access, the encoded video data output from the first processing chain 140 is stored in a continuous area of the file (shown as group 1 162). The encoded video data output from the second processing chain 150 is stored in another area 164 of the file 160. When it is desired to reproduce only a reduced-size video sequence during reproduction, only the encoded video data of group 1 162 may be extracted and decoded. On the other hand, when it is desired to decode a full-size video sequence, both groups 162 and 164 may be extracted and decoded. That is, according to the present invention, real-time image reproduction can be performed in various sizes.

  In one embodiment, all frames may be encoded as intra (intraframe) encoded frames (“I frames”). In intra coding, one block of video data in a frame is encoded from video data of other blocks in the frame. That is, the video content of the predetermined block X is encoded based on the data of the block adjacent to the block X in the horizontal or vertical direction. For example, the technique described in Chapter 7.4.3 of the MPEG-4 standard can be applied to the encoder 100. Information technology-Coding of audio-visual objects-Part 2: Visual Amenity 1: Visual extensions, ISO / IEC JTC 1 / SC 29 / WG 11 N 3056 For simplicity of explanation, it is assumed that such processing is included in the transformation block 110.

  In one embodiment, run length encoder 142 performs a modified zigzag scan on the small block pair. FIG. 3 shows a zigzag scan for a typical pair of 4 × 4 blocks in coefficient data. In one embodiment, the scan starts from the origin of the first block and advances the position of the coefficients in the first block according to the conventional zigzag scan direction. Scans in the block are performed perpendicular to the main diagonal of the block until the last coefficient is reached (compare with Fig. 2 (a)). The position of the last coefficient is generally located on the main diagonal of the subblock as far as possible from the origin. The scan direction jumps from this final position of the first block to the corresponding position of the second block, labeled “initial position” in FIG. The scanning direction of the coefficient position in the second block is the reverse direction. Scanning is performed in a direction perpendicular to the main diagonal of the block from the initial position of the second block to the origin. That is, the scan of the present embodiment processes two small blocks for each repeated scan.

  The embodiment shown in FIG. 3 is particularly efficient in coding small block coefficients. Traditional zigzag scans that scan only a single block because small blocks typically contain fewer coefficients than full-size blocks (eg, 16 coefficients for 64 coefficients) Then, the zero run tends to be relatively small and becomes inefficient. In contrast, by combining the conventional zigzag scan of the first block and the reverse zigzag scan of the second block, a longer zero run is likely to occur. In smooth image data, the coefficients near the final position of the first block and the initial position of the second block are often zero. A run (scan) that “jumps” from the first block to the second block causes these multiple zeros to be captured.

  FIG. 4 is a block diagram of an encoding chain 400 according to another embodiment of the present invention. The encoding chain 400 can be used as the first or second encoding chain shown in FIG. FIG. 4 shows a multimode run-length encoding system 410, which includes a run-length encoder 420 and a selector 430. The selector 430 has a pair of inputs. The first input is connected to the input to the encoding chain and the second input is connected to the output of the run-length encoding system 410. An input of run length encoder 420 is also connected to an input of run length encoding system 410. Based on a selection signal (denoted as “RLE selection” in FIG. 4), selector 430 passes data from one input selected from two inputs for further processing by the variable length encoder ( Forward).

  In other embodiments, the encoding algorithm applied to the data may be dynamically changed based on various data patterns input to the variable length encoder 440. Variable length encoder 440 is shown as having an encoder 450 and a plurality of code allocation tables 460. Based on the input selection signal, the VLC selector 470 associates one of the allocation tables with the encoding unit 450. Here, the input selection signal can be dynamically changed, for example, for each block, for each tile, or for each frame. Also, the VLC selection signal can be determined based on the block type (for example, luminance data or color data to be held), the coding chain (chain 140 or 150) to which the variable length encoder belongs, or the tile to which the block belongs. It is. Furthermore, if the blocks are encoded with several VLC allocation tables, the compression efficiency can be compared between the VLC allocation tables. The encoding result with the highest compression efficiency can be used and stored in the storage device.

  Even when a lossless encoding algorithm is used for encoding, the variable-length encoder 500 may encode input data by direct calculation or by using a lookup table.

  In other embodiments, the converter 110 may use a wavelet transform. Multi-resolution wavelet decomposition is one of the most efficient techniques for encoding image information. The encoded image information is expressed as a linear combination of locally supported wavelets. An example of wavelet support is shown in FIG. Wavelet decomposition yields DC coefficients (denoted as “DC”) and several coefficient bands corresponding to the horizontal, vertical, and diagonal wavelets of a given block. In FIG. 5, four coefficient bands (bands 1 to 4) are shown. The first band contains very low frequency wavelet coefficients. Each wavelet in band 2 divides the block area into quarters. Therefore, there are 4 horizontal wavelet coefficients, 4 vertical wavelet coefficients, and 4 diagonal wavelet coefficients. Wavelet coefficient In each further band, the coefficient represents a wavelet having a quarter of the corresponding wavelet region in the previous coefficient band. In FIG. 5, wavelet support is shown as image information at five different granularities in the block (bands 0-4, where band 0 is a DC coefficient).

  The principles of the present invention can be applied to wavelet-based encoders. In such an embodiment, the converter 110 may generate wavelet coefficients, and the generated wavelet coefficients may be truncated by the quantizer 120. The splitter 130 transfers the first set of band wavelet coefficients to the first encoding chain 140 and the remaining band wavelet coefficients to the second encoding chain 150. If the system generates a reduced size representation of 25% of the full size image, the splitter forwards the coefficients in bands 0-3 to the first encoding chain.

  FIG. 6 is a block diagram of a video decoder 600 according to an embodiment of the present invention. The decoder 600 can perform the reverse of the encoding process performed by the encoder 100 of FIG. Accordingly, the decoder comprises an inverse transformer 610, an inverse quantizer 620, a multiplexer 630, a pair of processing chains 640, 650, each having run length decoders 642, 652 and variable length decoders 644, 654, and storage. A device 660 will be provided. As described above, the encoded video data may exist in two separate groups in a file in the storage device. The first group includes encoded data of small blocks, and only this data needs to be decoded when it is desired to obtain a reduced-size copy of the source video image. The second group includes the encoded data of the remaining blocks, and a full-size image is generated by decoding together with the encoded data of the small blocks.

  In operation, one or both processing chains 640, 650 are active. When only the first processing chain 640 is active, small block encoded video data is retrieved from the storage device, variable length decoded 644 and run length decoded 642. When the second processing chain 650 is active, the encoded video data of the remaining blocks are also retrieved from the storage device, variable length decoded 654, and run length decoded 652. When multiplexer 630 receives data from both processing chains 640, 650, it combines the received data into a format for further processing at inverse quantizer 620 and inverse transformer 610.

The inverse quantizer 620 multiplies the input coefficient value by the quantization parameter (q _p ) applied in the quantizer 120 of FIG. 1 to obtain a transformed value. The inverse transformer 610 performs the inverse process of the transform (for example, discrete cosine transform coding or wavelet transform coding) applied by the converter 110 in FIG. 1 to restore the pixel value. Thus, the converter outputs a block of pixel data, and the output block can be reconstructed and displayed with other blocks. Here, the restored image data becomes a reduced size image when only the reduced size block is decoded, and becomes a full size image when all the encoded video data of the frame are decoded.

  In one embodiment, an identifier for the decoding mode to use is received when the inverse transformer 610 operates. When full size decoding is performed, the inverse converter 610 operates normally. However, when the reduced size decoding is performed, the inverse transformer 610 omits the inverse transformation calculation for only the coefficients in the remaining blocks, that is, the coefficients that are originally output from the second processing chain 650. This is because the chain 650 becomes inactive by the selection signal, so there is no point in performing the inverse transformation calculation. By omitting the calculation in this way, the throughput of the inverse converter 610 can be optimized.

  The inverse transformer 610 may be a wavelet transformer that generates pixel data from wavelet coefficients. In such an embodiment, if the inverse transformer operates only on small blocks, conventional wavelet decoding will produce a low resolution full size video because there are no high frequency coefficients. Thus, in one embodiment, the full size video is resized to the corresponding reduced size (eg, reduced to the original 25% size).

  As shown in Table 1 below, run-length encoded data is represented as a series of run and level pairs (consecutive). The run value is the number of consecutive coefficient scan positions where a zero coefficient is detected. The run value can be any value between zero and the maximum number of coefficient positions in the data unit (64 for 8 × 8 blocks, 16 for 4 × 4 block coded pairs). The level value is the value of the next non-zero coefficient detected in the scan direction. Therefore, in order to encode the coefficients of a single block or block pair, the run-length encoder generates at least one run, typically a sequence of runs and levels determined by the information content of the block. . Block encoding may end with either a run value or a level value.

  Typically, a run length encoder generates several run length spares.

  In general, if the last pair of run values can represent the remainder of a given block, the level will not continue as shown in Table 1, but the end of block symbol will follow.

  According to the embodiment of the present invention, by using an alternative “pattern” of run-length encoded data such as the pattern shown in Table 2, the data in Table 1 can be efficiently decoded.

  As described above, the n-th level symbol does not necessarily exist.

  FIG. 7 shows a reading method 700 in an embodiment of the present invention. According to this embodiment, the method reads and reads the first run value present in the run-length spare input pattern (box 710). Next, the method determines whether the current coefficient position specified from the sum of the run value read up to this point and the number of level values is at the final position or one position before the final position (box 720). . If it is not the final position or a position before the final position, the level value and the run value are read (box 730) and the process returns to box 720 for the next iteration. If the current coefficient position is the final position, the method 700 simply ends. If the current coefficient position is one position before the final position, the method reads the last level value (box 740) and ends.

  The reading process shown in FIG. 7 reads a pair of run and level according to the pattern of Table 2. This reading pattern is faster than the reading method based on the pattern shown in Table 1 because only one test is required for each repetition. Normally, run length decoding processes run values and level values as separate symbols, so a test must be performed to determine whether the final position has been reached each time each symbol is read. On the other hand, by reading the symbol pair as shown in Table 2, the method of FIG. 7 only needs to execute the test of half the number of times, and higher speed processing is possible.

  As described above, the encoding method of the present embodiment constitutes encoded video data according to a protocol that can decode a reduced-size image at an extremely high speed, and is particularly suitable for video editing applications. In the above description, an example of a 4 × 4 block displayed in a quarter size when decoded is shown. Of course, the principle of the present invention is not limited to this. The operation of the decoding system may be configured to provide at any desired ratio between reduced display and full size display. In the size ratios shown in Table 3, the splitter 130 can directly construct sub-blocks having the dimensions shown in the table from 8 × 8 source blocks.

  For example, if another size ratio such as 50% is desired, a further spatial size changing technique may be used for the decoded video data once output from the decoder (FIG. 6).

  Also, in order to provide further flexibility, video encoders and decoders comprise three or more encoding / decoding chains, each adapted to process video data corresponding to a predetermined frame size. Just do it. Using a video encoder as an example, the first encoding chain processes the 3 × 3 array of DCT coefficients obtained, and the second encoding chain processes the remaining coefficients up to a 5 × 5 coefficient array (3 The third encoded chain only needs to process the remaining coefficients that are not processed by the first two encoded chains. Similarly, a video decoder can comprise multiple encoding chains in a similar manner.

  The systems and techniques shown in the above embodiments can also be used in larger video coding systems that provide full coding services. One such system encodes video frames as multiple “tiles”. It is possible to divide the spatial area of the frame into different parts called tiles, and apply a predetermined encoding method used for the blocks in the tile not for each block but for each tile. For example, the selection of the VLC table (FIG. 4) is performed in common for all blocks in the very low tile. Also, encoding parameters such as quantization parameter selection are performed in common for all blocks in the tile. By applying the encoding method (encoding parameter) for each tile, the video data of the tile can be encoded / decoded independently of the video data belonging to another tile in the same frame. As a result, (1) individual processors can be allocated to tiles, and encoding / decoding can be performed in parallel. (2) Memory management and cache management can be performed by dividing video data of a frame into small tiles. (3) Encoding parameters can be selected independently for each tile (and optimization can be achieved), resulting in higher compression efficiency. (4) Interest in video sequences Several advantages are obtained, including that a tile can be decoded without decoding the entire frame to which the tile belongs. Therefore, a plurality of video encoders of FIG. 1 and video decoders of FIG. 6 are prepared, and a parallel processing system capable of independently decoding tiles can be configured.

  Various embodiments of the present invention have been specifically depicted and described herein. However, it should be understood that modifications and improvements of the invention are covered by the above-described techniques and fall within the scope of the appended claims without departing from the basic idea or intended scope of the invention.

1 is a block diagram of a video encoding system according to an embodiment of the present invention. It is a figure which shows the block classification based on embodiment of this invention. It is a figure which shows the scanning direction of the run length encoder based on embodiment of this invention. It is a figure which shows the block classification based on embodiment of this invention which performs an encoding by a wavelet coefficient. FIG. 3 is a diagram illustrating a variable length encoder according to an embodiment of the present invention. 1 is a block diagram of a video decoding system according to an embodiment of the present invention. 5 is a flowchart illustrating a run length reading method according to an embodiment of the present invention. It is a figure which shows the structural example of a computer.

Claims (21)

A converter that encodes pixel data of a plurality of blocks into transform coefficients;
A first variable length encoder that encodes a first sub-block including a preselected number of low frequency transform coefficients in each of the plurality of blocks;
A second variable length encoder that encodes a second sub-block including the remaining transform coefficients excluding the low-frequency transform coefficients in each of the plurality of blocks;
Equipped with a,
A first run length encoder for encoding the first sub-block;
A second run length encoder for encoding the second sub-block;
Further comprising
The first run-length encoder encodes the first sub-block pair together according to a predetermined scan pattern;
The predetermined scan pattern is:
Zigzag from the lowest frequency coefficient to the highest frequency coefficient in the first of the pair of sub-blocks,
Move to the highest frequency coefficient in the second of the pair of sub-blocks,
A scan pattern that zigzags from the highest frequency coefficient to the lowest frequency coefficient in the second of the pair of sub-blocks.
A video encoding system characterized by the above.   The video encoding system according to claim 1, wherein at least one of the first and second variable length encoders includes an encoder and a selectable variable length code allocation table.   The video encoding system of claim 1, wherein the first and second variable length encoders directly encode the first and second sub-blocks, respectively. Divide each frame of the input video into multiple pixel blocks,
For each block,
Encoding the block into a plurality of coefficients according to a predetermined transformation scheme;
Quantize the coefficients of the block according to predetermined quantization parameters;
Taking a subset of the coefficients from the block;
The extracted coefficient is variable length encoded, and the result of the encoding is stored in the first storage area,
A video encoding method characterized in that the remaining coefficients excluding the extracted coefficients are variable-length encoded, and the result of the encoding is stored in a second storage area ,
further,
Run-length encoding the extracted coefficients before variable length encoding;
Before the variable length coding, the remaining coefficients excluding the extracted coefficients are run-length coded.
It is characterized by
The run-length encoding of the extracted coefficients includes encoding together the pair of subsets of the extracted coefficients according to a predetermined scan pattern;
The predetermined scan pattern is:
Zigzag from the lowest frequency coefficient to the highest frequency coefficient in the first of the pair of subsets,
Move to the highest frequency coefficient in the second of the pair of subsets,
A scan pattern that zigzags from the highest frequency coefficient to the lowest frequency coefficient in the second of the pair of subsets
A video encoding method characterized by the above . 5. The video encoding method of claim 4 , wherein the variable length encoding includes directly encoding the extracted coefficient and the remaining coefficient excluding the extracted coefficient. 5. The video encoding method according to claim 4 , wherein the conversion method is a discrete cosine transform. 5. The video encoding method according to claim 4 , wherein the conversion method is wavelet decomposition. The video encoding method according to claim 4 , further comprising: predicting image data of the block from adjacent blocks for at least one block. 5. The video encoding method according to claim 4 , wherein each frame of the video sequence is encoded as an intra-encoded frame. A computer-readable storage medium storing encoded video data generated by a predetermined process, wherein the predetermined process includes:
Divide each frame of the input video into multiple pixel blocks,
For each block,
Encoding the block into a plurality of coefficients according to a predetermined transformation scheme;
Quantize the coefficients of the block according to predetermined quantization parameters;
Taking a subset of the coefficients from the block;
The extracted coefficient is variable length encoded, and the result of the encoding is stored in the first storage area,
Variable length encoding of the remaining coefficients excluding the extracted coefficients, and storing the result of the encoding in a second storage area;
The encoded data indicating the extracted coefficient is stored in a separate area in the computer-readable storage medium that is accessed independently from the encoded data indicating the remaining coefficient excluding the extracted coefficient. a computer-readable storage medium characterized by that,
The predetermined processing further includes
Run-length encoding the extracted coefficients before variable length encoding;
Before the variable length coding, the remaining coefficients excluding the extracted coefficients are run-length coded.
It is characterized by
The run-length encoding of the extracted coefficients includes encoding together the pair of subsets of the extracted coefficients according to a predetermined scan pattern;
The predetermined scan pattern is:
Zigzag from the lowest frequency coefficient to the highest frequency coefficient in the first of the pair of subsets,
Move to the highest frequency coefficient in the second of the pair of subsets,
A scan pattern that zigzags from the highest frequency coefficient to the lowest frequency coefficient in the second of the pair of subsets
A computer-readable storage medium . A first run-length decoder having first encoded video data as input;
A multiplexer connected to the output of the first run-length decoder and receiving the output;
An inverse converter connected to the output of the multiplexer;
With
The first run-length decoder state, and are not for decoding said first coded video data representing source video sequence of reduced size,
The first encoded video data is run-length encoded according to a predetermined scan pattern in units of a pair of low frequency sub-blocks,
The predetermined scan pattern is:
Zigzag from the lowest frequency coefficient to the highest frequency coefficient in the first of the pair of sub-blocks,
Move to the highest frequency coefficient in the second of the pair of sub-blocks,
Video decoder, wherein Ru said scan pattern der proceeding zigzag toward from the highest frequency coefficient in a second to the coefficient of the lowest frequency of said pair of sub-blocks. A second run-length decoder having the second encoded video data as input;
The video decoder according to claim 11 , wherein the second run-length decoder is for decoding the second encoded video data indicating a full-size source video sequence. A first variable length decoder to output the first encoded video data as an input, connected to the input to said input of said first run length decoder,
A second variable length decoder to the second encoded video data as an input, is connected to the input of the second run-length decoder outputs the input,
The video decoder of claim 12 , further comprising: 13. The video decoder of claim 12 , further comprising a control input indicating whether to disable the second run length decoder. The video decoder of claim 11 , wherein the inverse transformer is an inverse discrete cosine transformer. The video decoder of claim 11 , wherein the inverse transformer is an inverse wavelet transformer. Retrieving first encoded video data representing a reduced size image;
Decoding the first encoded video data to obtain a block of coefficients;
If the control signal indicates to perform full size decoding,
Retrieve the second encoded video data;
Decoding the second encoded video data to obtain a block of remaining coefficients that make up the coefficients;
Combining the coefficients and the remaining coefficients for each block;
A video decoding method characterized by generating pixel data from the combined coefficient block by inverse transformation ,
The first encoded video data is run-length encoded according to a predetermined scan pattern in units of a pair of low frequency sub-blocks,
The predetermined scan pattern is:
Zigzag from the lowest frequency coefficient to the highest frequency coefficient in the first of the pair of sub-blocks,
Move to the highest frequency coefficient in the second of the pair of sub-blocks,
A scan pattern that zigzags from the highest frequency coefficient to the lowest frequency coefficient in the second of the pair of sub-blocks.
A video decoding method characterized by the above . Generating the pixel data from the block of coefficients obtained by decoding the first encoded video data if the control signal indicates not to perform full size decoding; The video decoding method according to claim 17 , further comprising: The video decoding method according to claim 17 , wherein the inverse transform is an inverse discrete cosine transform. The video decoding method according to claim 17 , wherein the inverse transform is an inverse wavelet transform. The video decoding method according to claim 17 , wherein the decoding of the first encoded video data includes decoding variable length encoded data by direct calculation.

Images