CA2644379A1 - Visually lossless video data compression - Google Patents
Visually lossless video data compression Download PDFInfo
- Publication number
- CA2644379A1 CA2644379A1 CA 2644379 CA2644379A CA2644379A1 CA 2644379 A1 CA2644379 A1 CA 2644379A1 CA 2644379 CA2644379 CA 2644379 CA 2644379 A CA2644379 A CA 2644379A CA 2644379 A1 CA2644379 A1 CA 2644379A1
- Authority
- CA
- Canada
- Prior art keywords
- video data
- compression
- video
- compressed
- block
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
- Color Television Systems (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
An apparatus and method of performing YUV (or YCrCb) video compression prior to storage within a memory and decompression upon retrieval of the blocks from memory. Compression is performed utilizing a quantizer to compress video data to a desired overall compression ratio R, even though the luma and chroma contributions to compression can differ for each subblock, each preferably selected in response to texture estimation. Selections are made for each subblock to perform either linear or non-linear quantization during compression. Compression is performed without utilizing data from blocks outside of the block being compressed, wherein video blocks can be retrieved and decompressed in any desired order.
In one implementation, an encoder non-sequentially selects blocks from memory which are then decompressed and encoded. The compression may be beneficially utilized in a number of different video transmission and storage applications without departing from the invention.
In one implementation, an encoder non-sequentially selects blocks from memory which are then decompressed and encoded. The compression may be beneficially utilized in a number of different video transmission and storage applications without departing from the invention.
Description
VISUALLY LOSSLESS VIDEO DATA COMPRESSION
CROSS-REFERENCE TO RELATED APPLICATIONS
This application claims priority from U.S. provisional application serial number 61/015,174 filed on 12/19/2007, incorporated herein by reference in its entirety.
BACKGROUND OF THE INVENTION
Field of the Invention This invention pertains generally to video processing, and more particularly to the compression and decompression of video within video memory prior to encoding or output.
. , .. .. . . .... ...... . .~ :. ._.. .. . ...... . ... .. ..... .. . .......
........ .... .. . . .
Description of Related Art A typical video architecture is implemented to preprocess video data prior to storing it in a video frame memory typically accessed through an external bus. The video data is then retrieved from the video memory during an encoding process and/or for output. The amount of data transfer between the encoder and the video memory over the bus may be very large, leading to the need for large amounts of video memory, requiring enormous memory bus bandwidth, and consequently leading to high power consumption.
FIG. 1 shows a conventional video camera or still-picture camera architecture. The input video obtained by an input device, such as CCD or CMOS sensors, is then stored in an external memory after being preprocessed. In general preprocessing comprises noise reduction and video data conversion from RGB (Red, Green, Blue) to YUV (one luminance Y, two chrominance U + V). Other processing devices, such as a video encoder or a display device, then read the video data from the video memory.
Since the amount of video data is generally large, the required bus bandwidth for video data transfer through the external bus is significantly high.
In particular with regard to HD (High Definition) video application, the required bus bandwidth is enormous, wherein bandwidth cost as well as power consumption become very high making it difficult to implement low-cost image/video systems.
Another technical challenge in video camera systems is in regard to the required memory size. Most video devices are implemented in Soc (System on Chip). In general, the cost of an external memory (such as SDRAM) is generally higher than other devices. Therefore, a reduction of required memory size allows the overall system cost to be reduced.
Accordingly, a need exists for a method for reducing necessary storage bandwidth and memory for video data streams, and in particular when performed prior to the video encoding process. The present invention fulfills that need as well as others and overcomes the limitations of prior solutions.
BRIEF SUMMARY OF THE INVENTION
An apparatus and method is described for visually lossiess video data compression which utilizes a YUV Video Data Compression/Decompression Method. For the sake of simplicity, the term YUV video will be utilized herein, while it should be appreciated that YCrCb video, or other similar encodings are equally applicable to the teachings and claims recited herein. A video input is compressed by partitioning each frame into a set of compressed blocks having a given number of pixels for each component (e.g., N pixels for Y, M pixels for U, V). The compression blocks do not overlap other compression blocks and can thus be processed independently of one another without the need to reference information from other compression blocks, wherein the video data can be accessed randomly which is particularly well suited for use by a video encoder. As a result storage of the compression blocks in memory require less memory storage as well as bus bandwidth in view of the smaller size of the compressed blocks. Afterward, the compressed video data is retrieved in any desired order (e.g., non-sequentially) from memory and decompressed, such as by an encoder, and/or less preferably for output to a display. Decompression restores the data to its original format, wherein the device or application utilizing the decompressed data need not be aware that it was ever compressed.
Complexity estimations are preferably performed prior to compression so that optimum levels and modes of compression can be determined.
During compression, predictions are made to predict the current pixel value based on previous pixels, then subblock sizing is determined and subblocks partitioned. Decisions are then made as to the type of quantization (e.g., linear or non-linear), such as preferably for each subblock. For non-linear quantization, an aspect of the invention provides estimation of the quantization parameter (QP), wherein not all possible values for QP need be searched. Quantization is then performed based on the above determinations, and finally the subblocks of compressed data are packed into blocks. During compression, one aspect of the invention first performs the chroma compression, wherein information from chroma compression is made available and utilized during luma compression. Aspects of the present invention provide for the selection of different levels of compression in response to the complexity of the video data and the number of bits available.
Compressed luma and chroma are combined into a final compressed block.
Since the types and levels of quantization are determined in response to the complexity of the data, the compressed video data upon decompression (although not exactly matching the original video data) would still appear to a human viewer as visuaily lossless. Thus, the inventive apparatus and method operates with lowered memory requirements and reduced bandwidth while producing an output of a desired quality, up to that of being visually lossless.
The objectives of reducing bus bandwidth and/or memory size is given above as an example. The applications of the compression method is not limited to such specific applicafion benefits, but can be applied to provide benefits in any other applications that require video data compression, for example in transmission of compressed video data over a network prior to decompression, or in storing compressed video data to any other media device, prior to decompression, such as with hard disks or-memory.
The invention is amenable to being embodied in a number of ways, including but not limited to the following descriptions.
An implementation of the invention is a video memory apparatus configured for communicating video pixel data with one or more video modules, comprising: (a) a video compression module coupled through a signal bus to the video memory and configured for, (a)(i) compressing luminance and chrominance YUV input video data using a quantizer, without utilizing pixel data from pixels outside of each block, into compressed video data having a reduced number of bits per pixel, (a)(ii) storing the compressed video data within the video memory; and (b) a video decompression module configured for retrieving blocks of compressed video data in any order and decompressing the compressed video data stored in the video memory into decompressed video data which has the same format and approximates the original video data which was received and compressed by the video compression module, wherein said video decompression module configured to output the decompressed video data.
CROSS-REFERENCE TO RELATED APPLICATIONS
This application claims priority from U.S. provisional application serial number 61/015,174 filed on 12/19/2007, incorporated herein by reference in its entirety.
BACKGROUND OF THE INVENTION
Field of the Invention This invention pertains generally to video processing, and more particularly to the compression and decompression of video within video memory prior to encoding or output.
. , .. .. . . .... ...... . .~ :. ._.. .. . ...... . ... .. ..... .. . .......
........ .... .. . . .
Description of Related Art A typical video architecture is implemented to preprocess video data prior to storing it in a video frame memory typically accessed through an external bus. The video data is then retrieved from the video memory during an encoding process and/or for output. The amount of data transfer between the encoder and the video memory over the bus may be very large, leading to the need for large amounts of video memory, requiring enormous memory bus bandwidth, and consequently leading to high power consumption.
FIG. 1 shows a conventional video camera or still-picture camera architecture. The input video obtained by an input device, such as CCD or CMOS sensors, is then stored in an external memory after being preprocessed. In general preprocessing comprises noise reduction and video data conversion from RGB (Red, Green, Blue) to YUV (one luminance Y, two chrominance U + V). Other processing devices, such as a video encoder or a display device, then read the video data from the video memory.
Since the amount of video data is generally large, the required bus bandwidth for video data transfer through the external bus is significantly high.
In particular with regard to HD (High Definition) video application, the required bus bandwidth is enormous, wherein bandwidth cost as well as power consumption become very high making it difficult to implement low-cost image/video systems.
Another technical challenge in video camera systems is in regard to the required memory size. Most video devices are implemented in Soc (System on Chip). In general, the cost of an external memory (such as SDRAM) is generally higher than other devices. Therefore, a reduction of required memory size allows the overall system cost to be reduced.
Accordingly, a need exists for a method for reducing necessary storage bandwidth and memory for video data streams, and in particular when performed prior to the video encoding process. The present invention fulfills that need as well as others and overcomes the limitations of prior solutions.
BRIEF SUMMARY OF THE INVENTION
An apparatus and method is described for visually lossiess video data compression which utilizes a YUV Video Data Compression/Decompression Method. For the sake of simplicity, the term YUV video will be utilized herein, while it should be appreciated that YCrCb video, or other similar encodings are equally applicable to the teachings and claims recited herein. A video input is compressed by partitioning each frame into a set of compressed blocks having a given number of pixels for each component (e.g., N pixels for Y, M pixels for U, V). The compression blocks do not overlap other compression blocks and can thus be processed independently of one another without the need to reference information from other compression blocks, wherein the video data can be accessed randomly which is particularly well suited for use by a video encoder. As a result storage of the compression blocks in memory require less memory storage as well as bus bandwidth in view of the smaller size of the compressed blocks. Afterward, the compressed video data is retrieved in any desired order (e.g., non-sequentially) from memory and decompressed, such as by an encoder, and/or less preferably for output to a display. Decompression restores the data to its original format, wherein the device or application utilizing the decompressed data need not be aware that it was ever compressed.
Complexity estimations are preferably performed prior to compression so that optimum levels and modes of compression can be determined.
During compression, predictions are made to predict the current pixel value based on previous pixels, then subblock sizing is determined and subblocks partitioned. Decisions are then made as to the type of quantization (e.g., linear or non-linear), such as preferably for each subblock. For non-linear quantization, an aspect of the invention provides estimation of the quantization parameter (QP), wherein not all possible values for QP need be searched. Quantization is then performed based on the above determinations, and finally the subblocks of compressed data are packed into blocks. During compression, one aspect of the invention first performs the chroma compression, wherein information from chroma compression is made available and utilized during luma compression. Aspects of the present invention provide for the selection of different levels of compression in response to the complexity of the video data and the number of bits available.
Compressed luma and chroma are combined into a final compressed block.
Since the types and levels of quantization are determined in response to the complexity of the data, the compressed video data upon decompression (although not exactly matching the original video data) would still appear to a human viewer as visuaily lossless. Thus, the inventive apparatus and method operates with lowered memory requirements and reduced bandwidth while producing an output of a desired quality, up to that of being visually lossless.
The objectives of reducing bus bandwidth and/or memory size is given above as an example. The applications of the compression method is not limited to such specific applicafion benefits, but can be applied to provide benefits in any other applications that require video data compression, for example in transmission of compressed video data over a network prior to decompression, or in storing compressed video data to any other media device, prior to decompression, such as with hard disks or-memory.
The invention is amenable to being embodied in a number of ways, including but not limited to the following descriptions.
An implementation of the invention is a video memory apparatus configured for communicating video pixel data with one or more video modules, comprising: (a) a video compression module coupled through a signal bus to the video memory and configured for, (a)(i) compressing luminance and chrominance YUV input video data using a quantizer, without utilizing pixel data from pixels outside of each block, into compressed video data having a reduced number of bits per pixel, (a)(ii) storing the compressed video data within the video memory; and (b) a video decompression module configured for retrieving blocks of compressed video data in any order and decompressing the compressed video data stored in the video memory into decompressed video data which has the same format and approximates the original video data which was received and compressed by the video compression module, wherein said video decompression module configured to output the decompressed video data.
In the video apparatus, compression is preferably performed according to an overall compression ratio R, that controls the extent of block compression and which can be expressed ratiometrically, or in terms of the number of bits contained in the resultant block. In at least one preferred embodiment, the same or different compression levels are selected for luma and chroma while maintaining overall compression ratio R. In at least one preferred mode, unless chroma has a high complexity level, the extent of compression is selected by the apparatus to minimize bit use in chroma, while optimizing bit use in luma.
In at least one implementation of the video apparatus, texture complexity is estimated for the compression blocks prior to compression. The allocation of compression space to luma and chroma within each block, such as the number of bits to be allocated for luma and chroma within the compressed block, are determined in response to a process of estimating texture complexity.
In at least one implementation, luma and chroma information are combined into compressed video data blocks, which are preferably padded with padding bits to maintain a fixed size for the compressed blocks.
In at least one configuration of the apparatus, luma compression for a given block utilizes information determined during the chroma compression process for that same block. In at least one mode of the invention, video compression is performed utilizing non-linear quantization, such as within a combination of linear and non-linear quantization. In a preferred implementation, quantization step sizes of different precisions are utilized when performing non-linear quantization.
In at least one mode of the invention, pixel prediction is performed during compression. Pixel prediction starts at an initial reference pixel selected in the middle of the block from which a right and left prediction direction are defined that can be processed in parallel as desired. It will be appreciated that by selecting the middle pixel (or close to the middle) to start prediction, then the reference pixel remains the same for both right and left directions of prediction. Prediction of pixel values is performed by predicting current pixel values based on previous pixel values, and during which the prediction of the two directions is independent, thus allowing the prediction process for right and left directions to be performed in parallel to thus reduce required processing time.
In at least one implementation, subblock partitioning is performed according to a desired configuration. To determine subblock partitioning a subblock cost value is first computed for at least a portion of the possible subblock configurations, after which subblock configurations whose cost exceeds a given cost threshold, and/or the number of bits available within the compressed block, are discarded.
In at least one embodiment of the invention, the input video data for the apparatus is received from an image sensor, such as may be integrated within a video or still camera, or other imaging device. The formatting of the input video data is configured to include both luminance and chrominance information.
In at least one object of the invention, compression and decompression are performed toward reducing bus bandwidth and/or video memory requirements, such as utilized as a form of YUV video data preprocessing performed prior to encoding of the blocks which are retrieved from memory in an encoder specific manner which is typically non-sequential (e.g., not in block number order). Alternatively, or additionally, compression and decompression according to the invention may be practiced when transmitting compressed video data over a network prior to decompression, or for storing video data to a media device prior to decompression.
One embodiment of the invention is a video encoder apparatus for encoding YUV video data, comprising: (a) a video memory configured for communicating video pixel data with one or more video modules; (b) a video compression module coupled through a signal bus to the video memory and configured for, (b)(i) compressing luminance and chrominance YUV input video data using a quantizer into compressed video data having a reduced number of bits per pixel, without the need of referencing data from other blocks; (b)(ii) storing the compressed video data within the video memory; and (c) a video decompression module configured for block retrieval from video memory in any desired order and decompression of compressed video data into decompressed video data which has the same format and approximates the original video data which was received and compressed by the video compression module; and (d) an encoding module which non-sequentially selects video data blocks from the memory and which receives and encodes the decompressed video data.
One embodiment of the invention is a method of compressing and decompressing YUV video data, comprising: (a) compressing input video data, by a compression ratio R using a quantizer, into compressed video data blocks having a reduced number of bits in luma and/or chroma for each block of video data; (b) wherein the compression of the input video data is performed without utilizing data from pixels outside of the block being compressed; (c) storing the compressed video data into video memory; and (d) decompressing the compressed video data for any of the video data blocks, selected in sequential or non-sequential order, to generate a decompressed video data output. In one implementation, the compression and decompression is performed in combination with an encoding process for the video data wherein non-sequential blocks of video data from the memory are selected, decompressed according to the invention and received for encoding.
One embodiment of the invention is a method of compressing and decompressing YUV video data, comprising: (a) compressing input video data, by a compression ratio R using a quantizer, into compressed video data blocks having a reduced number of bits in luma and/or chroma for each block of video data; (b) the compression performed on each block of video without utilizing data from outside of each block being compressed; (c) selecting either linear or non-linear quantization for each subblock within a given block to which the compression is performed; (d) storing the compressed video data into video memory; and (e) decompressing the compressed video data for any of the video data blocks retrieved from video memory, selected in any order, to generate a decompressed video data output.
One embodiment of the invention is a method of compressing and decompressing YUV video data, comprising: (a) compressing input video data, by a compression ratio R using a quantizer, into compressed video data blocks having a reduced number of bits in luma and/or chroma for each block of video data; (b) the compression is performed on the video data blocks without utilizing data from blocks outside of the block being compressed; (c) estimating luma texture complexity and chroma texture complexity; (d) selecting the same or different compression levels within luma and chroma whiie maintaining overall compression ratio R; (e) selecting either linear or non-linear quantization for each subblock within a given block to which compression is to be performed in response to characteristics detected in the block; (f) wherein during compression, the luma compression process utilizes information from the chroma compression process for that same block; (g) storing the compressed video data into video memory; (h) retrieving blocks of video data from video memory in any desired order and at any time subsequent to said storing of the compressed video data; (i) decompressing the compressed video data for retrieved blocks to generate a decompressed video data output.
The present invention provides a number of beneficial aspects which can be implemented either separately or in any desired combination without departing from the present teachings.
An aspect of the invention comprises an apparatus and method for the compression and decompression of YUV video data blocks.
Another aspect of the invention is that each compressed block does not overlap with other compressed blocks, and each compression biock is compressed independently without referring to information from other blocks.
Another aspect of the invention is that, if the number of bits of the compressed information after compression is smaller than the target bits which comport with the desired compression ratio (R), then padding bits are used to extend the total output bits up to a desired number so that it remains fixed. Since the size of generated bits of each compression block is fixed, a compressed block at any position can be accessed and decompressed without referring to information in other blocks. Therefore, using the proposed method, access to video data and decompression of video blocks from a random position within the video can be readily performed, such as where an arbitrary area in a frame needs to be accessed for video processing, or encoding.
Another aspect of the invention is that luma data is compressed by using information determined as a result of chroma compression. Chroma is compressed first in this method according to the invention. After chroma compression, at least a portion of the information is transferred to a luma module in preparation for luma compression. The information bits can be retained for any desired period, or number of bits, after compression. Luma compression utilizes this information in compression. Finally, the compressed luma and chroma data is packed into one compressed block. The data size is guaranteed to be within the range specified by the target compression ratio.
Another aspect of the invention is that of first estimating, before compression, the texture complexity of luma and chroma compression blocks.
Each compression module (luma and chroma) can automatically provide different levels (or degrees) of compression (e.g., weak, medium, or strong).
The degree of compression is determined based on texture compiexity, wherein a number of bits is then allocated for luma and chroma fields within the compressed block.
Another aspect of the invention is the utilization of an overall compression ratio given by R which controls the overall extent to which the video blocks are compressed. The value of R is exemplified as a ratio (number of input bits I number of output bits) for a given block. A ratio R of two (2) indicates that the compressed blocks contain one-half the number of bits utilized by the original data. It should be appreciated, however, that compression can be expressed in any desired format, such as any form of ratiometric designation, or with regard to selecting a given number of bits per block of the video output.
Another aspect of the invention allows selecting the extent to which Y, U, and V are compressed while holding to an overall compression ratio R. It should be noted that it does not mean the compression ratio is fixed across Y, U and V for an individual compression block, as it possible to allocate bits based on different compression ratios between chroma and luma.
Another aspect of the invention is the process of automatically determining the best compression ratio between luma and chroma while maintaining an overall compression ratio R. In general terms, the human eye is more sensitive to signal noise in luma than signal noise in chroma.
Therefore, the preferred bit allocation strategy minimizes bit use for chroma, and maximizes bit use in luma. However, when chroma is found to have a certain level of complexity, in particular when it is found to be very complex, then at least one embodiment of the invention allows chroma to use a sufficient number of bits to prevent the appearance of visual artifacts. This maximization and minimization of bit use is determined in response to information obtained about luma and chroma texture complexity.
Another aspect of the invention selects different levels of compression.
For example, low, middle and high quality modes of compression can be defined. The low quality mode is based on utilizing a high compression ratio, while the high quality mode utilizes a low compression ratio. In one aspect of the invention, a determination is made of the final best compression mode. It should be noted that compression level can be defined both in luma and chroma compression, wherein the portion of the block used by luma and the portion used by chroma is determined.
Another aspect of the invention is that the decision to determine a best compression mode is made by information obtained about texture complexity and the number of available bits.
Another aspect of the invention starts pixel prediction from the middle pixel of a string of pixels instead of the first pixel, wherein other pixels are predicted in relation to that starting pixel. Two prediction directions are computed, wherein the required processing time can be reduced.
Another aspect of the invention is subblock partitioning in response to quantization and subblock configuration decisions. Overall compression ratio R is determined in response to the number of pixels N used for luma and the number of pixels M used for chroma compression (U and V, or Cr and Cb).
For a given (N, M) luma and chroma compression value, there exists a number of different subblock configurations. A subblock size decision module, or similar, determines an optimum subblock configuration under a given set of inputs and conditions.
Another aspect of the invention is for a cost to be computed for at least a portion of the different possible subblock configurations, (e.g., or all different subblock combinations) based on the given information, such as that received from Information from chroma, R, QP precision, and the like. By way of example, the cost can be an estimated in terms of the number of generated output bits. If the number of output bits for a given subblock configuration exceeds a threshold cost value, preferably given by the number of bits available, then that subblock configuration is discarded and the next possible subblock configuration is checked. In at least one implementation, if none of the possible configurations are avaiiabie, then the N pixels of the original compression block are used as a subblock.
Another aspect of the invention is the selecting of either linear or non-linear quantization during block compression.
Another aspect of the invention is estimation of a quantization parameter (QP) for use in non-linear quantization as described herein.
Further aspects of the invention will be brought out in the following portions of the specification, wherein the detailed description is for the purpose of fully disclosing preferred embodiments of the invention without placing limitations thereon.
BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWING(S) The invention will be more fully understood by reference to the following drawings which are for illustrative purposes only:
FIG. 1 is a block diagram of a conventional video or still camera architecture, showing storage of image data after preprocessing within a video memory.
FIG. 2 is a block diagram of a video or still camera architecture according to an aspect of the present invention, showing compression of image data prior to storage in video memory.
FIG. 3 is a block diagram of a block being compressed according to an aspect of the present invention.
FIG. 4 is a block diagram of block decompression according to an aspect of the present invention.
FIG. 5 is a flow diagram of chroma and luma compression according to an aspect of the present invention, showing information from chroma compression utilized in the luma compression process.
FIG. 6 is a flow diagram of overall compression according to an aspect of the present invention.
FIG. 7 is a flow diagram of supporting different compression modes according to an aspect of the present invention.
FIG. 8 is a flow diagram of making compression level decisions according to an aspect of the present invention, showing estimation of texture complexity and analysis in preparation of deciding on a compression level and performing the compression method.
FIG. 9 is a flow diagram of a compression method according to an aspect of the present invention, showing either linear or non-linear Q being performed for a block in response to a quantization decision.
FIG. 10 is a pixel diagram of pixel prediction according to an aspect of the present invention, showing selection of a reference pixel in the middle of the pixels.
FIG. 11 is a pixel diagram of right and left prediction directions in response to selection of a reference pixel as shown in FIG. 10.
FIG. 12 is a block diagram of selecting a subblock size according to an aspect of the present invention.
FIG. 13 is a flow diagram of determining subblock size according to an aspect of the present invention.
FIG. 14 is a flow diagram of non-linear quantization in response to estimated QP value and residue data according to an aspect of the present invention.
FIG. 15 is a flow diagram of QP estimation according to an aspect of i ..
the present invention, shown performed in response to determining residue data.
FIG. 16 is a block diagram of selecting linear or non-linear quantization according to an aspect of the present invention.
FIG. 17 is a flow diagram of deciding on linear or non-linear quantization according to an aspect of the present invention.
DETAILED DESCRIPTION OF THE INVENTION
Referring more specifically to the drawings, for illustrative purposes the present invention is embodied in the apparatus generally shown in FIG. 2 through FIG. 17. It will be appreciated that the apparatus may vary as to configuration and as to details of the parts, and that the method may vary as to the specific steps and sequence, without departing from the basic concepts as disclosed herein.
YUV Video Data Compression/Decompression Method FIG. 2 illustrates an embodiment 10 of the present invention for performing compression and decompression of YUV (or YCrCb) video data.
Compression may be utilized to render any number of benefits associated with a given application, for example in order to reduce bus bandwidth and video memory requirements.
The apparatus and method of the present invention receives input from video device 12, processes the data 14, and then compresses the video data 16 before storing it, such as over a bus 18, into an external video memory 20.
A video encoder 24 or display device 28 retrieves the compressed video from video memory and decompresses it 22, 26 prior to utilizing the video data.
Since the video data is compressed prior to storage in the video memory, the required bus bandwidth is much lower than the original, while the necessary size of the video memory (e.g., SDRAM) is similarly reduced.
FIG. 3 illustrates an example embodiment 30 of a YUV, or similar block, being compressed according to the present invention. The input video data format of compression block 32 of the invenfive compression method is YUV
(or YCrCb). During the compression process, a frame (that in this example consists of video data of Y, Cr, and Cb) is partitioned to a set of compression blocks. A compression block is defined as N pixels for Luminance (Y) 34 and M pixels each for chrominance (Cr, Cb) 26, 38 as shown in the figure.
Therefore, the total number of pixels in each compression block is (N+2M). If B bits are utilized to represent a pixel, then the total number of bits will be (N+2M)xB bits.
The compression biock does not overlap with other compression blocks, and each compression block is compressed independently without referring to information in other blocks.
Once a compression ratio, R is received for input, one implementation of compression method 40 generates fixed- size compressed blocks 42. The number of the generated bits is computed by:
Total Output Bits = (N+2M) * B / R.
For example, if R is 2, the total generated output bits will be half of the original number of bits of YUV.
If the number of bits of the compressed information after compression is smaller than the target bits specified by R, then padding bits are preferably inserted to maintain a fixed number of total output bits at all times. Since the size of generated bits of each compression block is fixed, a compressed block at any position can be retrieved and decompressed without referring to information in other blocks. Therefore, using the inventive method, access to the video data at a random position (as a compression block unit) is possible.
This is necessary in applications where an arbitrary area in a frame needs to be accessed for other video processing.
FIG. 4 depicts an embodiment 50 of the decompression process. The decompression method accepts a compression block 52 which is decompressed 54 to a compression block 56 having retumed to the original YUV resolution, such as having N pixels for Y 58, M pixels for U 60, and M
pixels for V. The decompressed YUV data can also be utilized in other devices for video processing. Since the dimension of video data after decompression is the same as the original, other processing devices will not be able to recognize, based on video formatting, any difference introduced by the compression algorithm.
... . ... .. ... . . ~. .. _ .. . ... .., . . . . . .... . .. . ...._ . .. ..
.......,.. .. . . .
FIG. 5 illustrates an embodiment 70 of block compression and the relation between luma and chroma compression modules. In at least one implementation, luma compression is performed in response to information received during compression of chroma for the same block. It will be seen that chroma (UV) data 72 is received first for chroma compression 74, wherein information from chroma compression, lnformation from chroma, is transferred and utilized in combination with luma (Y) data 76 to perform luma compression 78. It is thus preferable that chroma compression be performed first according to this inventive method. It should be appreciated that the information from chroma can be retained for any desired time, or number of bits following compression, such as for use during lUma compression. Finally, the compressed luma and chroma data is packed, as represented by block 80, into one compressed block 82. According to at least one preferred implementation, the data size is guaranteed to be within the range specified by the target compression ratio.
FIG. 6 illustrates an embodiment 90 of the overall inventive compression method. The receipt of a chroma compression block 92 and luma compression block 94 are shown. Prior to compression, the texture complexity of chroma is estimated 96 and the texture complexity of luma is estimated 100 (no specific order implied). Chroma compression 98 and luma compression 102 are performed with data received from both luma and chroma complexity estimation, while luma compression additionally receives information from chroma compression. Compressed luma and chroma data are then received and packaged 104 and a compressed block 106 generated.
Returning to consider the compression process, it will be appreciated that each compression module (luma and chroma) in the present invention may provide different levels (degrees) of compression (e.g., weak, medium, strong). According to at least one implementation of the invention, the degree of compression is preferably selected in response to the level of texture complexity.
The compression ratio represented by variable R controls the compression ratio of Y, U, and V compression blocks. It should be noted that it is not inferred from this that Y, U and V are subject to a fixed compression of R; because Y, U and V (or Y, Cr and Cb) can each be separately compressed to any desired ratio, insofar as the resultant compression for the block follows the overall compression ratio R. Therefore, it is possible to allocate bits based on different compression ratios between chroma and luma.
In one implementation of the invention, the embodied method attempts to optimize (i.e., find the best compression ratio within the confines of the technique and available information) compression ratio between luma and chroma while maintaining the overall compression ratio at R. This aspect of the invention takes into account that the human eye is generally more .
sensitive to noise in luma than to noise within chroma. Therefore, in the preferred bit allocation strategy, chroma utilizes a minimal number of bits so that bit use in luma can be maximized. However, in certain cases such as where chroma is found to be very complex, then a mode of the invention allows chroma to utilize a sufficient number of bits so to prevent visual artifacts.
In at least one implementation, the ratio of bit allocations is selected on the basis of a determination of luma and chroma texture complexity. The texture complexity in FIG. 6 can be obtained, for example, computing the average of residues of the compression block.
FIG. 7 illustrates an embodiment 110 of an inventive method to support different compression quality modes depending on compression level. For example, a compression block 112 can be subject to, low 114, medium 116, and high quality 118 compression modes according to the invention. The low quality mode arises from use of a high compression ratio, while the high quality video mode results from the use of a low compression ratio. In one preferred implementation, a process 120 is executed for selecting a final best compression mode from which a final compressed block 122 results. It should be recognized that the compression level can be defined both in luma and chroma compression modes.
One example of imptementing the different levels (modes) of compression is through utilizing bit precision of output quantization. For low compression modes, higher precisions can be utilized, while lower precisions can be utilized with high compression modes. Depending on the desired compression ratio, the quantization precisions can be predetermined at different levels. As shown in FIG. 7, the determination of the best compression mode is generated by the system in response to information on texture complexity and available bits.
FIG. 8 illustrates an example embodiment 130 of making compression level decisions according 'to the present invention. Lurna (Y) data 132 and chroma (UV) data 134 are received respectively for luma texture complexity estimation 136 and chroma texture complexity estimation 138. These complexity estimates are described in the following sections. Luma and chroma texture estimates are then utilized by a texture analyzer 140, the output from which is utilized in making compression level decisions in block 142 and then performing the compression method as per block 144 with a compressed chroma data output 146 and output of information from chroma 148. The compression method is described in FIG. 5 while the texture analysis and compression level decisions are described below.
Luma Texture Complexity Estimation The following pseudo code illustrates, by way of example and not limitation, the estimation of luma texture complexity. For each subblock, luma texture complexity is analyzed and the complexity is estimated as follows.
index x: index of pixel position index bp: index of subblock position Luma_Texture_Complexity (LTC) This value indicates the complexity of the current luma compression block.
Three different levels of complexity; weak, medium and strong can be defined.
The detection method:
For each subblock, compute Residual[x] = Current pixel[x] - predicted pixel[x];
For each subblock, find maximum of Residual[x] within subblock, say max Residual[bp];
Compute average_residue of max_Residue[bp] over whole compression block;
If average_residue < Threshoki_1:
The whole luma_block is considered as 'weak complex'.
If average_residue > Threshold_2:
Compute the number of subblocks that max Residue[bp] > Threshold_3;
If the computed number < Threshold_4;
Then, whole luma_block is considered as'medium complex block'.
Else, the whole luma block is considered as 'strong complex block'.
Chroma Texture Complexity Estimation The following pseudo code illustrates, by way of example and not limitation, estimation of chroma texture complexity. For each subblock, chroma texture complexity is analyzed and the complexity is estimated as follows.
Chroma Texture Complexity (CTC) The detection method:
For each subblock, compute Residual[x] = Current pixel[x] - predicted pixel[x];
If any Residual[x] < Threshold_1:
If the residue value at the subblock boundary is larger than Thresho{d_2;
Then, the subblock is considered as 'low complex'.
Else, the 4x1 chroma subblock is considered as 'high_complex' Compression Level Decision By way of example, and not limitation, the compression level decision can be implemented according to the pseudo-code below.
If (CTC == high) If(LTC ==low) Chroma_Compression_Level = LOW_COMPRESSION_MODE
Luma_Compression_Levei = HIGH COMPRESSION_MODE
Else if(LTC = medium) Chroma_Compression_Level = HIGH_COMPRESSION_MODE
Luma_Compression_Level = MIDDLE_COMPRESSION_MODE
Else (LTC==high) Chroma_Compression_Level = LOW_COMPRESSION_MODE
Luma_Compression_Level = HIGH_COMPRESSION_MODE
Else if (CTC == low) If(LTC ==low) Chroma_Compression_Level = LOW_COMPRESSION_MODE
Luma_Compression_Level = HIGH COMPRESSION_MODE
Else ff(LTC == medium) Chroma_Compression_Leve! = MIDDLE_COMPRESSION_MODE
Luma_Compression_Level = MIDDLE_COMPRESSION_MODE
Else (LTC=-high) Chroma_Compression_Level = HIGH_COMPRESSION_MODE
Luma_Compression_Level = MIDDLE_COMPRESSION_MODE
Compression Method In Each Level FIG. 9 illustrates an example embodiment 150 of an inventive compression method. A compression block 152 is received by a prediction process 154 which predicts the current pixel value based on previous pixels.
Once the predicted value is obtained, the residue value (difference between the current pixel and predicted pixel) is computed in the prediction process 154. A subblock size decision process 156 determines an optimal subblock size based on texture complexity and available bits budget. Once the subblock size is decided in 156, the set of residue data is partitioned at block 158 into smaller size arrays referred to as subblocks. A Quanfization process 160 is applied to each subblock of residue data to reduce the output bits.
Therefore, each subblock can be quantized using a unique QP (Quantization Parameter) value. The figure shows that the decision is made to perform a linear quantization 164, or estimate a quantization parameter (QP) 162 and then perform a non-linear quantization 166. In either case the subblock results are packed 168 to yield a final compressed block 170.
The prediction implies the current pixel is predicted from previous pixels. The predicted pixel is utilized for computing a 'residue' between the actual current pixel value and the predicted value. In the inventive method, the residue will be quantized and coded in later stages.
Prediction of x[n] = F(x[n-1], x[n-2], ... ) Residue of x[n] = abs(x[n] - Prediction of x[n]) FIG. 10 and FIG. 11 illustrate mechanisms for initial pixel prediction 190 and for supporting two prediction directions 210, respectively. It will be appreciated that the pixel data set may be of any size, with the example data set being shown for simplicity by way of example and not limitation. For simplicity of explanation, a one dimensional data set will be described, although the mechanisms can also be applied to two-dimensional data.
Within these figures, the number in the box indicates the position of the pixels in the compression block.
It should be noted in FIG. 10 that prediction 190 is started from the middle pixel 194 of data 192 instead of the first pixel. The pixel at the approximate middle position (e.g., position 16) 194 is set as the initial reference pixel, with all other pixels predicted starting from this pixel. As shown in FIG. 11 the mechanism 210 supports two prediction directions which are given as a right direction 212 and a left direction 214. Since the reference pixel (i.e., pixel at position 16) does not change, the prediction of each of these two directions is independent. Therefore, it is possible to process the prediction in parallel for the right and left directions, whereby the required processing time can be substantially reduced.
A compression block is partitioned to subblocks. The purpose of subblock partitions is to allocate independent QP for each subblock. Since lower QP can be used to minimize the noise caused by quantization, the overall quality can be improved if divided into a number of smaller subblocks.
However, if the number of subblocks is large, the overhead for the QP value increases.
Subblock Partitioning For Quantization And Subblock Configuration Decision Given (N, M) pixels, there are many possible subblock configurations which can be utilized. A'subblock size decision' module determines the best subblock configuration in response to the received parameters.
One example of subblock partitioning is based on the information obtained after chroma compression. In one aspect of the present invention, the chroma compression is carried out first wherein information from that process is made available for use during luma compression. For example, the number of total bits used in chroma compression and the chroma texture complexity can be considered as information from chroma.
FIG. 12 illustrates an embodiment 230 for making a subblock size decision within a subblock size decision module 232 in response to receipt of information from chroma, R, and QPprecision. Output of module 232 comprises a best subblock size, such as 2x2, 4x4, 8x8 and so forth.
FIG. 13 illustrates one possible implementation 250 of a `subblock size decision' module. The decision module receives 252 information from chroma, R, and QP_precision for each possible subblock configuration. A cost is computed for each configuration (254, 260, 266) based on the given information. The cost value can be expressed as an estimated number of generated output bits. A decision ladder (256, 262) is shown wherein if the number of output bits is larger than that available, then that particular configuration is discarded and a check is performed on the next possible configuration. Alternative outputs are here exemplified as 2x2 subblock 258, 4x4 subblock 264, and so forth up to a no partition (N subblock) output 266. It will be seen that when none of the possible configurations are acceptable, then the N pixels of the original compression block are used 266 as a subblock with no partition 268.
It should be appreciated that the inventive method is configured to support two quantization processes, linear and non-linear quantization.
1. Quantization Parameter Estimation In Non-linear Quantization Given the set of possible QP values, the system could check all of the QP values for a proper fit. However, the computational complexity of an exhaustive search for QP is not efficient and has a high implementation cost.
FIG. 14 illustrates an embodiment 270 of a reduced complexity method to estimate QP value without searching all possible QP values. Residue data 272 is shown received by QP estimator 274. Both the QP value determined from estimation and the residue data are received by a non-linear Q process 276.
FIG. 15 illustrates an embodiment 290 of QP estimation shown in detail.
The inventive QP estimator uses original pixel values with subblock residue data 292 utilized in finding a maximum residue value 294 in the subblocks.
The maximum value is then utilized in determining quantization decision level 296. In other words, from the maximum residue value a maximum decision level can be found from which the best QP value can be determined 298.
Table 1 shows one example of mapping maximum decision level and QP
value.
In order to decide the QP value of two-bit Q non-linear quantization, the maximum of residual data using original data within 4x1 block is used (QP
estimation). In Table 1 the mapping between the estimated QP and max residual data within subblock are seen.
2. Linear/N on-I i near Quantization Decision In this aspect of the present invention each subblock can be quantized n response to the selection of linear or non-linear quantization according to the inventive method. For example, in response to large residue values for a subblock the QP value also tends to be large, wherein significant quantization noise is introduced resulting in the creation of unwanted visual artifacts.
FIG. 16 illustrates an embodiment 310 of selecting non-linear or linear quantization, in response to the content of video data, in order to prevent visual artifacts from arising such as noise. In the figure, the example parameters that are shown being utilized in the selection are depicted as subblock residues 312 and original pixel data 314. However, any other parameters available at this stage can be used as well, within the quantization decision 316 to arrive at a final quantization method decision 318.
FIG. 17 illustrates an embodiment 330 of a'non-linear/linear quantization decision' module. Data, such as subblock residues 332 and original subblock pixel data values 334 are received within an edge detection module 336. It should be appreciated that edge detection 336 can be implemented in any number of alternative ways using original pixels or residues, or similar values. For example, if the maximum residue in a subblock is larger than a threshold, it can be considered that a strong edge exists. If no strong edges are detected, as per step 338, then a decision is made to use non-linear quantization 346. Othennrise a cost is estimated 340 to use linear quantization, which is then compared at 342 with the available number of bits to yield a decision to use linear quantization 344. If the cost is greater than the available bits, nonlinear quantization 346 is selected.
Although the description above contains many details, these should not be construed as limiting the scope of the invention but as merely providing illustrations of some of the presently preferred embodiments of this invention.
Therefore, it will be appreciated that the scope of the present invention fully encompasses other embodiments which may become obvious to those skilled in the art, and that the scope of the present invention is accordingly to be limited by nothing other than the appended claims, in which reference to an element in the singular is not intended to mean "one and only one" unless explicitly so stated, but rather "one or more." All structural and functional equivalents to the elements of the above-described preferred embodiment that are known to those of ordinary skill in the art are expressly incorporated herein by reference and are intended to be encompassed by the present claims. Moreover, it is not necessary for a device or method to address each and every problem sought to be solved by the present invention, for it to be encompassed by the present claims. Furthermore, no element, component, or method step in the present disclosure is intended to be dedicated to the public regardless of whether the element, component, or method step is explicitly recited in the claims. No claim element herein is to be construed under the provisions of 35 U.S.C. 112, sixth paragraph, unless the element is expressly recited using the phrase "means for."
Table 1 Example Mapping of Maximum Decision Level And QP Value DL DO Dl D2 D3 D4 D5 D6 D7 MRR r<4 4!5r<4 8_r<16 165r<32 32<_r<48 48_r<64 645r<128 128:5r<256 E.QP 0 1 2 3 4 5 6 7 DL - Decision Level; MRR - Maximum Residue Range; E. QP - Estimated QP
In at least one implementation of the video apparatus, texture complexity is estimated for the compression blocks prior to compression. The allocation of compression space to luma and chroma within each block, such as the number of bits to be allocated for luma and chroma within the compressed block, are determined in response to a process of estimating texture complexity.
In at least one implementation, luma and chroma information are combined into compressed video data blocks, which are preferably padded with padding bits to maintain a fixed size for the compressed blocks.
In at least one configuration of the apparatus, luma compression for a given block utilizes information determined during the chroma compression process for that same block. In at least one mode of the invention, video compression is performed utilizing non-linear quantization, such as within a combination of linear and non-linear quantization. In a preferred implementation, quantization step sizes of different precisions are utilized when performing non-linear quantization.
In at least one mode of the invention, pixel prediction is performed during compression. Pixel prediction starts at an initial reference pixel selected in the middle of the block from which a right and left prediction direction are defined that can be processed in parallel as desired. It will be appreciated that by selecting the middle pixel (or close to the middle) to start prediction, then the reference pixel remains the same for both right and left directions of prediction. Prediction of pixel values is performed by predicting current pixel values based on previous pixel values, and during which the prediction of the two directions is independent, thus allowing the prediction process for right and left directions to be performed in parallel to thus reduce required processing time.
In at least one implementation, subblock partitioning is performed according to a desired configuration. To determine subblock partitioning a subblock cost value is first computed for at least a portion of the possible subblock configurations, after which subblock configurations whose cost exceeds a given cost threshold, and/or the number of bits available within the compressed block, are discarded.
In at least one embodiment of the invention, the input video data for the apparatus is received from an image sensor, such as may be integrated within a video or still camera, or other imaging device. The formatting of the input video data is configured to include both luminance and chrominance information.
In at least one object of the invention, compression and decompression are performed toward reducing bus bandwidth and/or video memory requirements, such as utilized as a form of YUV video data preprocessing performed prior to encoding of the blocks which are retrieved from memory in an encoder specific manner which is typically non-sequential (e.g., not in block number order). Alternatively, or additionally, compression and decompression according to the invention may be practiced when transmitting compressed video data over a network prior to decompression, or for storing video data to a media device prior to decompression.
One embodiment of the invention is a video encoder apparatus for encoding YUV video data, comprising: (a) a video memory configured for communicating video pixel data with one or more video modules; (b) a video compression module coupled through a signal bus to the video memory and configured for, (b)(i) compressing luminance and chrominance YUV input video data using a quantizer into compressed video data having a reduced number of bits per pixel, without the need of referencing data from other blocks; (b)(ii) storing the compressed video data within the video memory; and (c) a video decompression module configured for block retrieval from video memory in any desired order and decompression of compressed video data into decompressed video data which has the same format and approximates the original video data which was received and compressed by the video compression module; and (d) an encoding module which non-sequentially selects video data blocks from the memory and which receives and encodes the decompressed video data.
One embodiment of the invention is a method of compressing and decompressing YUV video data, comprising: (a) compressing input video data, by a compression ratio R using a quantizer, into compressed video data blocks having a reduced number of bits in luma and/or chroma for each block of video data; (b) wherein the compression of the input video data is performed without utilizing data from pixels outside of the block being compressed; (c) storing the compressed video data into video memory; and (d) decompressing the compressed video data for any of the video data blocks, selected in sequential or non-sequential order, to generate a decompressed video data output. In one implementation, the compression and decompression is performed in combination with an encoding process for the video data wherein non-sequential blocks of video data from the memory are selected, decompressed according to the invention and received for encoding.
One embodiment of the invention is a method of compressing and decompressing YUV video data, comprising: (a) compressing input video data, by a compression ratio R using a quantizer, into compressed video data blocks having a reduced number of bits in luma and/or chroma for each block of video data; (b) the compression performed on each block of video without utilizing data from outside of each block being compressed; (c) selecting either linear or non-linear quantization for each subblock within a given block to which the compression is performed; (d) storing the compressed video data into video memory; and (e) decompressing the compressed video data for any of the video data blocks retrieved from video memory, selected in any order, to generate a decompressed video data output.
One embodiment of the invention is a method of compressing and decompressing YUV video data, comprising: (a) compressing input video data, by a compression ratio R using a quantizer, into compressed video data blocks having a reduced number of bits in luma and/or chroma for each block of video data; (b) the compression is performed on the video data blocks without utilizing data from blocks outside of the block being compressed; (c) estimating luma texture complexity and chroma texture complexity; (d) selecting the same or different compression levels within luma and chroma whiie maintaining overall compression ratio R; (e) selecting either linear or non-linear quantization for each subblock within a given block to which compression is to be performed in response to characteristics detected in the block; (f) wherein during compression, the luma compression process utilizes information from the chroma compression process for that same block; (g) storing the compressed video data into video memory; (h) retrieving blocks of video data from video memory in any desired order and at any time subsequent to said storing of the compressed video data; (i) decompressing the compressed video data for retrieved blocks to generate a decompressed video data output.
The present invention provides a number of beneficial aspects which can be implemented either separately or in any desired combination without departing from the present teachings.
An aspect of the invention comprises an apparatus and method for the compression and decompression of YUV video data blocks.
Another aspect of the invention is that each compressed block does not overlap with other compressed blocks, and each compression biock is compressed independently without referring to information from other blocks.
Another aspect of the invention is that, if the number of bits of the compressed information after compression is smaller than the target bits which comport with the desired compression ratio (R), then padding bits are used to extend the total output bits up to a desired number so that it remains fixed. Since the size of generated bits of each compression block is fixed, a compressed block at any position can be accessed and decompressed without referring to information in other blocks. Therefore, using the proposed method, access to video data and decompression of video blocks from a random position within the video can be readily performed, such as where an arbitrary area in a frame needs to be accessed for video processing, or encoding.
Another aspect of the invention is that luma data is compressed by using information determined as a result of chroma compression. Chroma is compressed first in this method according to the invention. After chroma compression, at least a portion of the information is transferred to a luma module in preparation for luma compression. The information bits can be retained for any desired period, or number of bits, after compression. Luma compression utilizes this information in compression. Finally, the compressed luma and chroma data is packed into one compressed block. The data size is guaranteed to be within the range specified by the target compression ratio.
Another aspect of the invention is that of first estimating, before compression, the texture complexity of luma and chroma compression blocks.
Each compression module (luma and chroma) can automatically provide different levels (or degrees) of compression (e.g., weak, medium, or strong).
The degree of compression is determined based on texture compiexity, wherein a number of bits is then allocated for luma and chroma fields within the compressed block.
Another aspect of the invention is the utilization of an overall compression ratio given by R which controls the overall extent to which the video blocks are compressed. The value of R is exemplified as a ratio (number of input bits I number of output bits) for a given block. A ratio R of two (2) indicates that the compressed blocks contain one-half the number of bits utilized by the original data. It should be appreciated, however, that compression can be expressed in any desired format, such as any form of ratiometric designation, or with regard to selecting a given number of bits per block of the video output.
Another aspect of the invention allows selecting the extent to which Y, U, and V are compressed while holding to an overall compression ratio R. It should be noted that it does not mean the compression ratio is fixed across Y, U and V for an individual compression block, as it possible to allocate bits based on different compression ratios between chroma and luma.
Another aspect of the invention is the process of automatically determining the best compression ratio between luma and chroma while maintaining an overall compression ratio R. In general terms, the human eye is more sensitive to signal noise in luma than signal noise in chroma.
Therefore, the preferred bit allocation strategy minimizes bit use for chroma, and maximizes bit use in luma. However, when chroma is found to have a certain level of complexity, in particular when it is found to be very complex, then at least one embodiment of the invention allows chroma to use a sufficient number of bits to prevent the appearance of visual artifacts. This maximization and minimization of bit use is determined in response to information obtained about luma and chroma texture complexity.
Another aspect of the invention selects different levels of compression.
For example, low, middle and high quality modes of compression can be defined. The low quality mode is based on utilizing a high compression ratio, while the high quality mode utilizes a low compression ratio. In one aspect of the invention, a determination is made of the final best compression mode. It should be noted that compression level can be defined both in luma and chroma compression, wherein the portion of the block used by luma and the portion used by chroma is determined.
Another aspect of the invention is that the decision to determine a best compression mode is made by information obtained about texture complexity and the number of available bits.
Another aspect of the invention starts pixel prediction from the middle pixel of a string of pixels instead of the first pixel, wherein other pixels are predicted in relation to that starting pixel. Two prediction directions are computed, wherein the required processing time can be reduced.
Another aspect of the invention is subblock partitioning in response to quantization and subblock configuration decisions. Overall compression ratio R is determined in response to the number of pixels N used for luma and the number of pixels M used for chroma compression (U and V, or Cr and Cb).
For a given (N, M) luma and chroma compression value, there exists a number of different subblock configurations. A subblock size decision module, or similar, determines an optimum subblock configuration under a given set of inputs and conditions.
Another aspect of the invention is for a cost to be computed for at least a portion of the different possible subblock configurations, (e.g., or all different subblock combinations) based on the given information, such as that received from Information from chroma, R, QP precision, and the like. By way of example, the cost can be an estimated in terms of the number of generated output bits. If the number of output bits for a given subblock configuration exceeds a threshold cost value, preferably given by the number of bits available, then that subblock configuration is discarded and the next possible subblock configuration is checked. In at least one implementation, if none of the possible configurations are avaiiabie, then the N pixels of the original compression block are used as a subblock.
Another aspect of the invention is the selecting of either linear or non-linear quantization during block compression.
Another aspect of the invention is estimation of a quantization parameter (QP) for use in non-linear quantization as described herein.
Further aspects of the invention will be brought out in the following portions of the specification, wherein the detailed description is for the purpose of fully disclosing preferred embodiments of the invention without placing limitations thereon.
BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWING(S) The invention will be more fully understood by reference to the following drawings which are for illustrative purposes only:
FIG. 1 is a block diagram of a conventional video or still camera architecture, showing storage of image data after preprocessing within a video memory.
FIG. 2 is a block diagram of a video or still camera architecture according to an aspect of the present invention, showing compression of image data prior to storage in video memory.
FIG. 3 is a block diagram of a block being compressed according to an aspect of the present invention.
FIG. 4 is a block diagram of block decompression according to an aspect of the present invention.
FIG. 5 is a flow diagram of chroma and luma compression according to an aspect of the present invention, showing information from chroma compression utilized in the luma compression process.
FIG. 6 is a flow diagram of overall compression according to an aspect of the present invention.
FIG. 7 is a flow diagram of supporting different compression modes according to an aspect of the present invention.
FIG. 8 is a flow diagram of making compression level decisions according to an aspect of the present invention, showing estimation of texture complexity and analysis in preparation of deciding on a compression level and performing the compression method.
FIG. 9 is a flow diagram of a compression method according to an aspect of the present invention, showing either linear or non-linear Q being performed for a block in response to a quantization decision.
FIG. 10 is a pixel diagram of pixel prediction according to an aspect of the present invention, showing selection of a reference pixel in the middle of the pixels.
FIG. 11 is a pixel diagram of right and left prediction directions in response to selection of a reference pixel as shown in FIG. 10.
FIG. 12 is a block diagram of selecting a subblock size according to an aspect of the present invention.
FIG. 13 is a flow diagram of determining subblock size according to an aspect of the present invention.
FIG. 14 is a flow diagram of non-linear quantization in response to estimated QP value and residue data according to an aspect of the present invention.
FIG. 15 is a flow diagram of QP estimation according to an aspect of i ..
the present invention, shown performed in response to determining residue data.
FIG. 16 is a block diagram of selecting linear or non-linear quantization according to an aspect of the present invention.
FIG. 17 is a flow diagram of deciding on linear or non-linear quantization according to an aspect of the present invention.
DETAILED DESCRIPTION OF THE INVENTION
Referring more specifically to the drawings, for illustrative purposes the present invention is embodied in the apparatus generally shown in FIG. 2 through FIG. 17. It will be appreciated that the apparatus may vary as to configuration and as to details of the parts, and that the method may vary as to the specific steps and sequence, without departing from the basic concepts as disclosed herein.
YUV Video Data Compression/Decompression Method FIG. 2 illustrates an embodiment 10 of the present invention for performing compression and decompression of YUV (or YCrCb) video data.
Compression may be utilized to render any number of benefits associated with a given application, for example in order to reduce bus bandwidth and video memory requirements.
The apparatus and method of the present invention receives input from video device 12, processes the data 14, and then compresses the video data 16 before storing it, such as over a bus 18, into an external video memory 20.
A video encoder 24 or display device 28 retrieves the compressed video from video memory and decompresses it 22, 26 prior to utilizing the video data.
Since the video data is compressed prior to storage in the video memory, the required bus bandwidth is much lower than the original, while the necessary size of the video memory (e.g., SDRAM) is similarly reduced.
FIG. 3 illustrates an example embodiment 30 of a YUV, or similar block, being compressed according to the present invention. The input video data format of compression block 32 of the invenfive compression method is YUV
(or YCrCb). During the compression process, a frame (that in this example consists of video data of Y, Cr, and Cb) is partitioned to a set of compression blocks. A compression block is defined as N pixels for Luminance (Y) 34 and M pixels each for chrominance (Cr, Cb) 26, 38 as shown in the figure.
Therefore, the total number of pixels in each compression block is (N+2M). If B bits are utilized to represent a pixel, then the total number of bits will be (N+2M)xB bits.
The compression biock does not overlap with other compression blocks, and each compression block is compressed independently without referring to information in other blocks.
Once a compression ratio, R is received for input, one implementation of compression method 40 generates fixed- size compressed blocks 42. The number of the generated bits is computed by:
Total Output Bits = (N+2M) * B / R.
For example, if R is 2, the total generated output bits will be half of the original number of bits of YUV.
If the number of bits of the compressed information after compression is smaller than the target bits specified by R, then padding bits are preferably inserted to maintain a fixed number of total output bits at all times. Since the size of generated bits of each compression block is fixed, a compressed block at any position can be retrieved and decompressed without referring to information in other blocks. Therefore, using the inventive method, access to the video data at a random position (as a compression block unit) is possible.
This is necessary in applications where an arbitrary area in a frame needs to be accessed for other video processing.
FIG. 4 depicts an embodiment 50 of the decompression process. The decompression method accepts a compression block 52 which is decompressed 54 to a compression block 56 having retumed to the original YUV resolution, such as having N pixels for Y 58, M pixels for U 60, and M
pixels for V. The decompressed YUV data can also be utilized in other devices for video processing. Since the dimension of video data after decompression is the same as the original, other processing devices will not be able to recognize, based on video formatting, any difference introduced by the compression algorithm.
... . ... .. ... . . ~. .. _ .. . ... .., . . . . . .... . .. . ...._ . .. ..
.......,.. .. . . .
FIG. 5 illustrates an embodiment 70 of block compression and the relation between luma and chroma compression modules. In at least one implementation, luma compression is performed in response to information received during compression of chroma for the same block. It will be seen that chroma (UV) data 72 is received first for chroma compression 74, wherein information from chroma compression, lnformation from chroma, is transferred and utilized in combination with luma (Y) data 76 to perform luma compression 78. It is thus preferable that chroma compression be performed first according to this inventive method. It should be appreciated that the information from chroma can be retained for any desired time, or number of bits following compression, such as for use during lUma compression. Finally, the compressed luma and chroma data is packed, as represented by block 80, into one compressed block 82. According to at least one preferred implementation, the data size is guaranteed to be within the range specified by the target compression ratio.
FIG. 6 illustrates an embodiment 90 of the overall inventive compression method. The receipt of a chroma compression block 92 and luma compression block 94 are shown. Prior to compression, the texture complexity of chroma is estimated 96 and the texture complexity of luma is estimated 100 (no specific order implied). Chroma compression 98 and luma compression 102 are performed with data received from both luma and chroma complexity estimation, while luma compression additionally receives information from chroma compression. Compressed luma and chroma data are then received and packaged 104 and a compressed block 106 generated.
Returning to consider the compression process, it will be appreciated that each compression module (luma and chroma) in the present invention may provide different levels (degrees) of compression (e.g., weak, medium, strong). According to at least one implementation of the invention, the degree of compression is preferably selected in response to the level of texture complexity.
The compression ratio represented by variable R controls the compression ratio of Y, U, and V compression blocks. It should be noted that it is not inferred from this that Y, U and V are subject to a fixed compression of R; because Y, U and V (or Y, Cr and Cb) can each be separately compressed to any desired ratio, insofar as the resultant compression for the block follows the overall compression ratio R. Therefore, it is possible to allocate bits based on different compression ratios between chroma and luma.
In one implementation of the invention, the embodied method attempts to optimize (i.e., find the best compression ratio within the confines of the technique and available information) compression ratio between luma and chroma while maintaining the overall compression ratio at R. This aspect of the invention takes into account that the human eye is generally more .
sensitive to noise in luma than to noise within chroma. Therefore, in the preferred bit allocation strategy, chroma utilizes a minimal number of bits so that bit use in luma can be maximized. However, in certain cases such as where chroma is found to be very complex, then a mode of the invention allows chroma to utilize a sufficient number of bits so to prevent visual artifacts.
In at least one implementation, the ratio of bit allocations is selected on the basis of a determination of luma and chroma texture complexity. The texture complexity in FIG. 6 can be obtained, for example, computing the average of residues of the compression block.
FIG. 7 illustrates an embodiment 110 of an inventive method to support different compression quality modes depending on compression level. For example, a compression block 112 can be subject to, low 114, medium 116, and high quality 118 compression modes according to the invention. The low quality mode arises from use of a high compression ratio, while the high quality video mode results from the use of a low compression ratio. In one preferred implementation, a process 120 is executed for selecting a final best compression mode from which a final compressed block 122 results. It should be recognized that the compression level can be defined both in luma and chroma compression modes.
One example of imptementing the different levels (modes) of compression is through utilizing bit precision of output quantization. For low compression modes, higher precisions can be utilized, while lower precisions can be utilized with high compression modes. Depending on the desired compression ratio, the quantization precisions can be predetermined at different levels. As shown in FIG. 7, the determination of the best compression mode is generated by the system in response to information on texture complexity and available bits.
FIG. 8 illustrates an example embodiment 130 of making compression level decisions according 'to the present invention. Lurna (Y) data 132 and chroma (UV) data 134 are received respectively for luma texture complexity estimation 136 and chroma texture complexity estimation 138. These complexity estimates are described in the following sections. Luma and chroma texture estimates are then utilized by a texture analyzer 140, the output from which is utilized in making compression level decisions in block 142 and then performing the compression method as per block 144 with a compressed chroma data output 146 and output of information from chroma 148. The compression method is described in FIG. 5 while the texture analysis and compression level decisions are described below.
Luma Texture Complexity Estimation The following pseudo code illustrates, by way of example and not limitation, the estimation of luma texture complexity. For each subblock, luma texture complexity is analyzed and the complexity is estimated as follows.
index x: index of pixel position index bp: index of subblock position Luma_Texture_Complexity (LTC) This value indicates the complexity of the current luma compression block.
Three different levels of complexity; weak, medium and strong can be defined.
The detection method:
For each subblock, compute Residual[x] = Current pixel[x] - predicted pixel[x];
For each subblock, find maximum of Residual[x] within subblock, say max Residual[bp];
Compute average_residue of max_Residue[bp] over whole compression block;
If average_residue < Threshoki_1:
The whole luma_block is considered as 'weak complex'.
If average_residue > Threshold_2:
Compute the number of subblocks that max Residue[bp] > Threshold_3;
If the computed number < Threshold_4;
Then, whole luma_block is considered as'medium complex block'.
Else, the whole luma block is considered as 'strong complex block'.
Chroma Texture Complexity Estimation The following pseudo code illustrates, by way of example and not limitation, estimation of chroma texture complexity. For each subblock, chroma texture complexity is analyzed and the complexity is estimated as follows.
Chroma Texture Complexity (CTC) The detection method:
For each subblock, compute Residual[x] = Current pixel[x] - predicted pixel[x];
If any Residual[x] < Threshold_1:
If the residue value at the subblock boundary is larger than Thresho{d_2;
Then, the subblock is considered as 'low complex'.
Else, the 4x1 chroma subblock is considered as 'high_complex' Compression Level Decision By way of example, and not limitation, the compression level decision can be implemented according to the pseudo-code below.
If (CTC == high) If(LTC ==low) Chroma_Compression_Level = LOW_COMPRESSION_MODE
Luma_Compression_Levei = HIGH COMPRESSION_MODE
Else if(LTC = medium) Chroma_Compression_Level = HIGH_COMPRESSION_MODE
Luma_Compression_Level = MIDDLE_COMPRESSION_MODE
Else (LTC==high) Chroma_Compression_Level = LOW_COMPRESSION_MODE
Luma_Compression_Level = HIGH_COMPRESSION_MODE
Else if (CTC == low) If(LTC ==low) Chroma_Compression_Level = LOW_COMPRESSION_MODE
Luma_Compression_Level = HIGH COMPRESSION_MODE
Else ff(LTC == medium) Chroma_Compression_Leve! = MIDDLE_COMPRESSION_MODE
Luma_Compression_Level = MIDDLE_COMPRESSION_MODE
Else (LTC=-high) Chroma_Compression_Level = HIGH_COMPRESSION_MODE
Luma_Compression_Level = MIDDLE_COMPRESSION_MODE
Compression Method In Each Level FIG. 9 illustrates an example embodiment 150 of an inventive compression method. A compression block 152 is received by a prediction process 154 which predicts the current pixel value based on previous pixels.
Once the predicted value is obtained, the residue value (difference between the current pixel and predicted pixel) is computed in the prediction process 154. A subblock size decision process 156 determines an optimal subblock size based on texture complexity and available bits budget. Once the subblock size is decided in 156, the set of residue data is partitioned at block 158 into smaller size arrays referred to as subblocks. A Quanfization process 160 is applied to each subblock of residue data to reduce the output bits.
Therefore, each subblock can be quantized using a unique QP (Quantization Parameter) value. The figure shows that the decision is made to perform a linear quantization 164, or estimate a quantization parameter (QP) 162 and then perform a non-linear quantization 166. In either case the subblock results are packed 168 to yield a final compressed block 170.
The prediction implies the current pixel is predicted from previous pixels. The predicted pixel is utilized for computing a 'residue' between the actual current pixel value and the predicted value. In the inventive method, the residue will be quantized and coded in later stages.
Prediction of x[n] = F(x[n-1], x[n-2], ... ) Residue of x[n] = abs(x[n] - Prediction of x[n]) FIG. 10 and FIG. 11 illustrate mechanisms for initial pixel prediction 190 and for supporting two prediction directions 210, respectively. It will be appreciated that the pixel data set may be of any size, with the example data set being shown for simplicity by way of example and not limitation. For simplicity of explanation, a one dimensional data set will be described, although the mechanisms can also be applied to two-dimensional data.
Within these figures, the number in the box indicates the position of the pixels in the compression block.
It should be noted in FIG. 10 that prediction 190 is started from the middle pixel 194 of data 192 instead of the first pixel. The pixel at the approximate middle position (e.g., position 16) 194 is set as the initial reference pixel, with all other pixels predicted starting from this pixel. As shown in FIG. 11 the mechanism 210 supports two prediction directions which are given as a right direction 212 and a left direction 214. Since the reference pixel (i.e., pixel at position 16) does not change, the prediction of each of these two directions is independent. Therefore, it is possible to process the prediction in parallel for the right and left directions, whereby the required processing time can be substantially reduced.
A compression block is partitioned to subblocks. The purpose of subblock partitions is to allocate independent QP for each subblock. Since lower QP can be used to minimize the noise caused by quantization, the overall quality can be improved if divided into a number of smaller subblocks.
However, if the number of subblocks is large, the overhead for the QP value increases.
Subblock Partitioning For Quantization And Subblock Configuration Decision Given (N, M) pixels, there are many possible subblock configurations which can be utilized. A'subblock size decision' module determines the best subblock configuration in response to the received parameters.
One example of subblock partitioning is based on the information obtained after chroma compression. In one aspect of the present invention, the chroma compression is carried out first wherein information from that process is made available for use during luma compression. For example, the number of total bits used in chroma compression and the chroma texture complexity can be considered as information from chroma.
FIG. 12 illustrates an embodiment 230 for making a subblock size decision within a subblock size decision module 232 in response to receipt of information from chroma, R, and QPprecision. Output of module 232 comprises a best subblock size, such as 2x2, 4x4, 8x8 and so forth.
FIG. 13 illustrates one possible implementation 250 of a `subblock size decision' module. The decision module receives 252 information from chroma, R, and QP_precision for each possible subblock configuration. A cost is computed for each configuration (254, 260, 266) based on the given information. The cost value can be expressed as an estimated number of generated output bits. A decision ladder (256, 262) is shown wherein if the number of output bits is larger than that available, then that particular configuration is discarded and a check is performed on the next possible configuration. Alternative outputs are here exemplified as 2x2 subblock 258, 4x4 subblock 264, and so forth up to a no partition (N subblock) output 266. It will be seen that when none of the possible configurations are acceptable, then the N pixels of the original compression block are used 266 as a subblock with no partition 268.
It should be appreciated that the inventive method is configured to support two quantization processes, linear and non-linear quantization.
1. Quantization Parameter Estimation In Non-linear Quantization Given the set of possible QP values, the system could check all of the QP values for a proper fit. However, the computational complexity of an exhaustive search for QP is not efficient and has a high implementation cost.
FIG. 14 illustrates an embodiment 270 of a reduced complexity method to estimate QP value without searching all possible QP values. Residue data 272 is shown received by QP estimator 274. Both the QP value determined from estimation and the residue data are received by a non-linear Q process 276.
FIG. 15 illustrates an embodiment 290 of QP estimation shown in detail.
The inventive QP estimator uses original pixel values with subblock residue data 292 utilized in finding a maximum residue value 294 in the subblocks.
The maximum value is then utilized in determining quantization decision level 296. In other words, from the maximum residue value a maximum decision level can be found from which the best QP value can be determined 298.
Table 1 shows one example of mapping maximum decision level and QP
value.
In order to decide the QP value of two-bit Q non-linear quantization, the maximum of residual data using original data within 4x1 block is used (QP
estimation). In Table 1 the mapping between the estimated QP and max residual data within subblock are seen.
2. Linear/N on-I i near Quantization Decision In this aspect of the present invention each subblock can be quantized n response to the selection of linear or non-linear quantization according to the inventive method. For example, in response to large residue values for a subblock the QP value also tends to be large, wherein significant quantization noise is introduced resulting in the creation of unwanted visual artifacts.
FIG. 16 illustrates an embodiment 310 of selecting non-linear or linear quantization, in response to the content of video data, in order to prevent visual artifacts from arising such as noise. In the figure, the example parameters that are shown being utilized in the selection are depicted as subblock residues 312 and original pixel data 314. However, any other parameters available at this stage can be used as well, within the quantization decision 316 to arrive at a final quantization method decision 318.
FIG. 17 illustrates an embodiment 330 of a'non-linear/linear quantization decision' module. Data, such as subblock residues 332 and original subblock pixel data values 334 are received within an edge detection module 336. It should be appreciated that edge detection 336 can be implemented in any number of alternative ways using original pixels or residues, or similar values. For example, if the maximum residue in a subblock is larger than a threshold, it can be considered that a strong edge exists. If no strong edges are detected, as per step 338, then a decision is made to use non-linear quantization 346. Othennrise a cost is estimated 340 to use linear quantization, which is then compared at 342 with the available number of bits to yield a decision to use linear quantization 344. If the cost is greater than the available bits, nonlinear quantization 346 is selected.
Although the description above contains many details, these should not be construed as limiting the scope of the invention but as merely providing illustrations of some of the presently preferred embodiments of this invention.
Therefore, it will be appreciated that the scope of the present invention fully encompasses other embodiments which may become obvious to those skilled in the art, and that the scope of the present invention is accordingly to be limited by nothing other than the appended claims, in which reference to an element in the singular is not intended to mean "one and only one" unless explicitly so stated, but rather "one or more." All structural and functional equivalents to the elements of the above-described preferred embodiment that are known to those of ordinary skill in the art are expressly incorporated herein by reference and are intended to be encompassed by the present claims. Moreover, it is not necessary for a device or method to address each and every problem sought to be solved by the present invention, for it to be encompassed by the present claims. Furthermore, no element, component, or method step in the present disclosure is intended to be dedicated to the public regardless of whether the element, component, or method step is explicitly recited in the claims. No claim element herein is to be construed under the provisions of 35 U.S.C. 112, sixth paragraph, unless the element is expressly recited using the phrase "means for."
Table 1 Example Mapping of Maximum Decision Level And QP Value DL DO Dl D2 D3 D4 D5 D6 D7 MRR r<4 4!5r<4 8_r<16 165r<32 32<_r<48 48_r<64 645r<128 128:5r<256 E.QP 0 1 2 3 4 5 6 7 DL - Decision Level; MRR - Maximum Residue Range; E. QP - Estimated QP
Claims (25)
1. An apparatus for compressing and decompressing YUV video data, comprising:
a video memory configured for communicating video pixel data with one or more video modules;
a video compression module coupled through a signal bus to said video memory and configured for, compressing blocks of luminance and chrominance YUV input video data using a quantizer into compressed video data having a reduced number of bits per pixel, wherein block compression is performed without utilizing pixel data outside of each block, storing said compressed video data within said video memory; and a video decompression module configured for retrieving blocks of compressed video data in any order and for decompressing the compressed video data stored in said video memory into decompressed video data which has the same format and approximates the original video data which was received and compressed by said video compression module;
said video decompression module configured to output said decompressed video data.
a video memory configured for communicating video pixel data with one or more video modules;
a video compression module coupled through a signal bus to said video memory and configured for, compressing blocks of luminance and chrominance YUV input video data using a quantizer into compressed video data having a reduced number of bits per pixel, wherein block compression is performed without utilizing pixel data outside of each block, storing said compressed video data within said video memory; and a video decompression module configured for retrieving blocks of compressed video data in any order and for decompressing the compressed video data stored in said video memory into decompressed video data which has the same format and approximates the original video data which was received and compressed by said video compression module;
said video decompression module configured to output said decompressed video data.
2. An apparatus as recited in claim 1, wherein said compression is performed according to an overall compression ratio R, which controls the extent of block compression.
3. An apparatus as recited in claim 2, wherein said overall compression ratio R can be expressed ratiometrically, or in terms of the number of bits contained in the resultant block.
4. An apparatus as recited in claim 2, further comprising selecting the same or different compression levels for luma and chroma while maintaining overall compression ratio R.
5. An apparatus as recited in claim 4, wherein unless chroma has a high complexity level, the extent of said compression is selected to minimize bit use in chroma, and maximize bit use in luma.
6. An apparatus as recited in claim 1, further comprising estimating texture complexity of compression blocks prior to said compression.
7. An apparatus as recited in claim 6, wherein the number of bits to be allocated for luma and chroma within the compressed block are determined in response to estimating texture complexity.
8. An apparatus as recited in claim 1, further comprising padding the compressed video data with padding bits to maintain a fixed size for the compressed blocks.
9. An apparatus as recited in claim 1, wherein Luma compression for a given block utilizes information from Chroma compression of that same block.
10. An apparatus as recited in claim 1, further comprising:
performing pixel prediction during compression, said pixel prediction starting at an initial reference pixel selected in the middle of the block and defining a right and left prediction direction which can be processed in parallel as desired;
wherein the reference pixel remains the same for both right and left directions of prediction.
performing pixel prediction during compression, said pixel prediction starting at an initial reference pixel selected in the middle of the block and defining a right and left prediction direction which can be processed in parallel as desired;
wherein the reference pixel remains the same for both right and left directions of prediction.
11. An apparatus as recited in claim 10, wherein said prediction of pixel values predicts current pixel values based on previous pixel values.
12. An apparatus as recited in claim 10, wherein during said prediction of pixel values the prediction of the two directions is independent, whereby the prediction process for right and left directions is performed in parallel and the required processing time will be reduced to one-half.
13. An apparatus as recited in claim 1, further comprising subblock partitioning according to a desired configuration.
14. An apparatus as recited in claim 11, further comprising computing a subblock cost value for at least a portion of the possible subblock configurations and discarding subblock configurations whose cost exceeds a threshold, or the number of bits available.
15. An apparatus as recited in claim 1, wherein said input video data is received from an image sensor.
16. An apparatus as recited in claim 1, wherein said apparatus is integrated within a video or still camera.
17. An apparatus as recited in claim 1, wherein said input video data comprises a format having luminance and chrominance YUV information in the form Y, Cr and Cb.
18. An apparatus as recited in claim 1, wherein said compression and decompression is performed to reduce bus bandwidth and video memory requirements when encoding video data.
19. An apparatus as recited in claim 1, wherein said apparatus is configured for compressing and decompressing video data prior to video encoding, for transmitting compressed video data over a network prior to decompression, or for storing video data to a media device prior to decompression.
20. An apparatus as recited in claim 1, wherein:
said video compression is performed utilizing non-linear quantization; and different precisions of quantization step sizes are utilized when performing said non-linear quantization.
said video compression is performed utilizing non-linear quantization; and different precisions of quantization step sizes are utilized when performing said non-linear quantization.
21. A video encoder apparatus for encoding YUV video data, comprising:
a video memory configured with a signal bus configured for communicating video pixel data with one or more video modules;
a video compression module coupled through said signal bus to said video memory and configured for, compressing luminance and chrominance YUV input video data using a quantizer, without utilizing pixel data from pixels outside of each block, into compressed video data having a reduced number of bits per pixel, storing said compressed video data within said video memory; and a video decompression module configured for retrieval and decompressing of the compressed video data stored in said video memory into decompressed video data which has the same format and approximates the original video data which was received and compressed by said video compression module; and an encoding module configured for non-sequentially selecting which video data blocks are to be retrieved by said video decompression module and for encoding the decompressed video data output from said decompression module.
a video memory configured with a signal bus configured for communicating video pixel data with one or more video modules;
a video compression module coupled through said signal bus to said video memory and configured for, compressing luminance and chrominance YUV input video data using a quantizer, without utilizing pixel data from pixels outside of each block, into compressed video data having a reduced number of bits per pixel, storing said compressed video data within said video memory; and a video decompression module configured for retrieval and decompressing of the compressed video data stored in said video memory into decompressed video data which has the same format and approximates the original video data which was received and compressed by said video compression module; and an encoding module configured for non-sequentially selecting which video data blocks are to be retrieved by said video decompression module and for encoding the decompressed video data output from said decompression module.
22. A method of compressing and decompressing YUV video data, comprising:
compressing input video data to a compression ratio R using a quantizer into compressed video data blocks having a reduced number of bits in luma and/or chroma for each block of video data;
wherein said compressing of the input video data is performed without utilizing data from pixels outside of the block being compressed;
storing said compressed video data into video memory; and decompressing said compressed video data for any of the video data blocks, retrieved in any desired order from said video memory to generate a decompressed video data output.
compressing input video data to a compression ratio R using a quantizer into compressed video data blocks having a reduced number of bits in luma and/or chroma for each block of video data;
wherein said compressing of the input video data is performed without utilizing data from pixels outside of the block being compressed;
storing said compressed video data into video memory; and decompressing said compressed video data for any of the video data blocks, retrieved in any desired order from said video memory to generate a decompressed video data output.
23. A method as recited in claim 22, further comprising:
encoding said video data in response to non-sequentially selecting blocks of video data from said memory and receiving said decompressed video data.
encoding said video data in response to non-sequentially selecting blocks of video data from said memory and receiving said decompressed video data.
24. A method of compressing and decompressing YUV video data, comprising:
compressing input video data to a compression ratio R using a quantizer into compressed video data blocks having a reduced number of bits in luma and/or chroma for each block of video data;
said compression is performed on the video data without utilizing data from pixels outside of the block being compressed;
selecting either linear or non-linear quantization for each subblock within a given block to which said compression is to be performed;
storing said compressed video data into video memory; and decompressing said compressed video data for any of the video data blocks, retrieved in any desired order from said video memory, to generate a decompressed video data output.
compressing input video data to a compression ratio R using a quantizer into compressed video data blocks having a reduced number of bits in luma and/or chroma for each block of video data;
said compression is performed on the video data without utilizing data from pixels outside of the block being compressed;
selecting either linear or non-linear quantization for each subblock within a given block to which said compression is to be performed;
storing said compressed video data into video memory; and decompressing said compressed video data for any of the video data blocks, retrieved in any desired order from said video memory, to generate a decompressed video data output.
25. A method of compressing and decompressing YUV video data, comprising:
compressing input video data to a compression ratio R using a quantizer into compressed video data blocks having a reduced number of bits in luma and/or chroma for each block of video data;
said compression is performed on the video data blocks without utilizing pixel data from outside of the block being compressed;
estimating luma texture complexity and chroma texture complexity;
selecting individual compression levels for luma and chroma while maintaining overall compression ratio R;
selecting either linear or non-linear quantization for each subblock within a given block to which said compression is to be performed in response to block characteristics;
wherein during said compression, the luma compression process utilizes information from the chroma compression process for that same block;
storing said compressed video data into video memory;
retrieving blocks of said video data from said video memory in any desired order and at any time subsequent to said storing of said compressed video data; and decompressing said compressed video data for retrieved blocks to generate a decompressed video data output.
compressing input video data to a compression ratio R using a quantizer into compressed video data blocks having a reduced number of bits in luma and/or chroma for each block of video data;
said compression is performed on the video data blocks without utilizing pixel data from outside of the block being compressed;
estimating luma texture complexity and chroma texture complexity;
selecting individual compression levels for luma and chroma while maintaining overall compression ratio R;
selecting either linear or non-linear quantization for each subblock within a given block to which said compression is to be performed in response to block characteristics;
wherein during said compression, the luma compression process utilizes information from the chroma compression process for that same block;
storing said compressed video data into video memory;
retrieving blocks of said video data from said video memory in any desired order and at any time subsequent to said storing of said compressed video data; and decompressing said compressed video data for retrieved blocks to generate a decompressed video data output.
Applications Claiming Priority (4)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US1517407P | 2007-12-19 | 2007-12-19 | |
| US61/015,174 | 2007-12-19 | ||
| US12/103,419 | 2008-04-15 | ||
| US12/103,419 US8401071B2 (en) | 2007-12-19 | 2008-04-15 | Virtually lossless video data compression |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CA2644379A1 true CA2644379A1 (en) | 2009-06-19 |
| CA2644379C CA2644379C (en) | 2015-01-06 |
Family
ID=40788592
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CA 2644379 Expired - Fee Related CA2644379C (en) | 2007-12-19 | 2008-11-21 | Visually lossless video data compression |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN101472182B (en) |
| CA (1) | CA2644379C (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9083954B2 (en) | 2011-11-02 | 2015-07-14 | Huawei Technologies Co., Ltd. | Video processing method and system and related device |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102469306A (en) * | 2010-10-29 | 2012-05-23 | 华晶科技股份有限公司 | Image compression method |
| CN102868869B (en) * | 2011-07-06 | 2015-05-13 | 华晶科技股份有限公司 | Image compression method, decompression method and electronic device with fixed compression ratio |
| WO2013179101A1 (en) * | 2012-06-01 | 2013-12-05 | Freescale Semiconductor, Inc. | Method and apparatus for performing compression of image data |
| CN109672923B (en) * | 2018-12-17 | 2021-07-02 | 龙迅半导体（合肥）股份有限公司 | Data processing method and device |
| CN114501024B (en) * | 2022-04-02 | 2022-07-19 | 苏州浪潮智能科技有限公司 | Video compression system, method, computer readable storage medium and server |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6043845A (en) * | 1997-08-29 | 2000-03-28 | Logitech | Video capture and compression system and method for composite video |
| US6208350B1 (en) * | 1997-11-04 | 2001-03-27 | Philips Electronics North America Corporation | Methods and apparatus for processing DVD video |
| KR100366382B1 (en) * | 2001-02-12 | 2002-12-31 | (주) 멀티비아 | Apparatus and method for coding moving picture |
-
2008
- 2008-11-21 CA CA 2644379 patent/CA2644379C/en not_active Expired - Fee Related
- 2008-12-19 CN CN2008101861727A patent/CN101472182B/en not_active Expired - Fee Related
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9083954B2 (en) | 2011-11-02 | 2015-07-14 | Huawei Technologies Co., Ltd. | Video processing method and system and related device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101472182A (en) | 2009-07-01 |
| CA2644379C (en) | 2015-01-06 |
| CN101472182B (en) | 2013-06-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US8401071B2 (en) | Virtually lossless video data compression | |
| CN107211128B (en) | Adaptive chroma downsampling and color space conversion techniques | |
| US7460131B2 (en) | Methods and systems for processing image data for display on LC displays | |
| EP1610563A2 (en) | Selecting encoding types and predictive modes for encoding video data | |
| EP2697971B1 (en) | Multiple color channel multiple regression predictor | |
| KR101535784B1 (en) | Method and system for illumination compensation and transition for video coding and processing | |
| US6301392B1 (en) | Efficient methodology to select the quantization threshold parameters in a DWT-based image compression scheme in order to score a predefined minimum number of images into a fixed size secondary storage | |
| US7409099B1 (en) | Method of improved image/video compression via data re-ordering | |
| US20020021756A1 (en) | Video compression using adaptive selection of groups of frames, adaptive bit allocation, and adaptive replenishment | |
| US20090129466A1 (en) | Method, medium, and apparatus efficiently encoding and decoding moving image using image resolution adjustment | |
| WO2008020687A1 (en) | Image encoding/decoding method and apparatus | |
| CA2644379A1 (en) | Visually lossless video data compression | |
| US20110235715A1 (en) | Video coding system and circuit emphasizing visual perception | |
| JP2007503784A (en) | Hybrid video compression method | |
| EP1997317A1 (en) | Image encoding/decoding method and apparatus | |
| US6556718B1 (en) | Video pictures compression and coding with decision step for field/frame and motion vectors DCT | |
| CN108289223A (en) | It is a kind of to can be used for the method for compressing image in liquid crystal display over-driving device and device | |
| CA2832228A1 (en) | Method for efficient representation and processing of color pixel data in digital pathology images | |
| US20100183076A1 (en) | Encoding Images | |
| Lalithambigai et al. | A hybrid adaptive block based compressive sensing in video for IoMT applications | |
| US8428381B2 (en) | Image compression method with variable quantization parameter | |
| Wu et al. | Efficient rate-control system with three stages for JPEG2000 image coding | |
| WO2013038888A1 (en) | Image encoding device and image decoding device | |
| Park et al. | Adaptive lossless video compression | |
| US7787537B2 (en) | Method and apparatus for low-memory high-performance discrete cosine transform coefficient prediction |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| EEER | Examination request | ||
| MKLA | Lapsed |
Effective date: 20161121 |