Pulse or digital communications – Bandwidth reduction or expansion – Television or motion video signal
Reexamination Certificate
2001-05-16
2004-09-14
Kelley, Chris (Department: 2613)
Pulse or digital communications
Bandwidth reduction or expansion
Television or motion video signal
Reexamination Certificate
active
06792044
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to methods of and systems for processing video during compression, specifically MPEG-4 video compression, wherein enhancement layers are added to the base layer using activity-based frequency weighting methods in an adaptive procedure, to allow a more visually-sensitive component of a video frame to be encoded in a high priority of number of bit-planes, and thus to provide high visual quality at decompression time
BACKGROUND OF THE INVENTION
“MPEG” generally represents an evolving set of standards for video and audio compression developed by the Moving Picture Experts Group. The need for compression of motion video for digital transmission becomes apparent with even a cursory look at uncompressed-bitrates in contrast with bandwidths available. MPEG-1 was designed for coding progressive video at a transmission rate of about 1.5 million bits per second. It was designed specifically for Video-CD and CD-i media. MPEG-2 was designed for coding interlaced images at transmission rates above 4 million bits per second. The MPEG-2 standard is used for various applications, such as digital television (DTV) broadcasts, digital versatile disk (DVD) technology, and video storage systems. MPEG-4 is designed for very low-bit rate applications, using a more flexible coding standard to target internet video transmission and the wireless communications market.
The MPEG4 video compression standard allows content-based access or transmission of an arbitrarily-shaped video object plane (VOP) at various temporal and spatial resolutions. MPEG4 supports both object and quality scalability. The fine granularity scalability (“FGS”) is one type of scalable coding scheme that is adopted by the MPEG4 standard. The FGS encoding scheme allows an MPEG4 bitstream to be encoded in two layers: the base layer, which encodes each frame with a fixed lower bound bit-rate; and the enhancement layer, which encodes the difference between original picture and the reconstructed base-layer picture. The enhancement layer is encoded via a bitplane coding scheme, therefore enhancement layer bitstrearms are scalable in the sense that an arbitrary (fine grained) number of bit-planes of the enhancement-layer can be transmitted to the decoder depending on the transmission bandwidth. The FGS coding scheme has been finalized by MPEG4 version 4.
In the standardized FGS scheme, frequency weighting is a feature used for visual quality improvement. By giving different weights to the elements of each coding block, the enhancement layer residuals are weighted and encoded relative to their importance to the visual output quality.
The MPEG-4 decoder may decode only the base layer or the base layer and any subset of the FGS enhancement layer. This is useful when the decoding device is of limited or variant bandwidth and for storage purposes.
In some cases, the base layer alone is decoded, allowing for a less-detailed video image to be viewed. When the bandwidth between encoding and decoding is variant, or when the space for bitstream storage is limited, the base layer is decoded and as much of the FGS enhancement layer can be added on top of the base layer as bandwidth or storage space allows.
The MPEG-4 standard operates by first encoding a base layer of the scene being compressed. This base layer is a lower-quality, low-bandwidth, compressed image. The base layer is represented by a plurality of coding blocks, e.g., the discrete cosine transform-encoded (“DCT”) blocks. The FGS enhancement layer is represented by a plurality of residual blocks. Next, the FGS enhancement layer generates a bitstream in addition to the base layer bitstream. Depending on the bandwidth of the transmission channel and complexity of the decoder, a truncated bitstream of the FGS layer will necessarily be decoded.
In the MPEG-4 coding standard, two quality improvement methods are standardized for FGS enhancement layer encoding. These two methods are frequency weighting and selective enhancement. Herein, only the frequency weighting method is addressed.
The FGS enhancement layer is used to code the quantization residuals from the base layer, therefore the overall quality of the coded sequence is the.combination of the base layer information and the transmitted FGS enhancement layer information. In theory, the FGS method codes the residuals of the base layer without loss. However, it is often the case that only part of the FGS enhancement layer can go through the transmission channel and arrive at the decoder, due to a limited transmission bandwidth. When bandwidth variation occurs, the number of bits of the FGS enhancement layer transmitted from the encoder side to the decoder side are variant depending on the bandwidth at the moment of transmittal. Also, due to the nature of FGS enhancement layer coding, it can be placed in storage to any desired quality level. Hence, the visual quality of the transmitted/stored signal is heavily impacted by the amount of the FGS layer that is decoded.
To improve the visual quality of the output sequence, frequency weighting allows the weighting of the elements of the residual block unevenly before the bitplane coding (which is the method used for the FGS layer coding). Since certain frequency components are visually more important, they should be enhanced more (i.e. they should be coded with high accuracy by being given high frequency weights), thereby improving the subjective image quality.
Bit-Plane Shift
Objects being encoded by bit-plane encoding are ordered from most-significant bit (“MSB”) to least significant bit (“LSB”). A Bit-plane shift describes the operation of shifting the bitplanes corresponding to a particular value in a block by one or more bits towards the MSB. This has the effect of increasing, or boosting, the priority of the objects encoded, in this case the residual block.
Quantization Residuals
When the base-layer coefficients are encoded or “quantized, ” the quantization function has an associated loss. Thus, the accuracy of the quantized data depends on the quantization steps. Quantization residuals are left out as a non-encoded part for the base-layer and not recoverable at the base layer of the decoder side.
Fine Granularity
Fine granularity refers to a coding method where the video data is encoded in a progressive way (bit-plane by bit-plane), from MSB to LSB. Consequently the encoded bitstream can be truncated at any bit-plane level, while always ensuring the more significant data is more likely to be sent.
Frequency weighting (“FW”) uses a FW matrix to selectively re-weight the importance of each enhancement layer coefficient within each coding block, so that the significance of each coefficient for bitplane encoding is re-prioritized by the weighting matrix. Each element of the FW matrix indicates the number of bit-plane shifts of the corresponding FGS coefficient within the block. A bit-plane shift of one is equivalent to the multiplication of the FGS coefficient by a power of two. While MPEG-4 does standardize the FGS tool, it does not provide an appropriate FW matrix. The FW matrix definition is left as an encoder optimization parameter to be set by each manufacturer individually.
Using a DCT based codec as an example, for an 8×8 DCT block, the DC coefficient and the lower frequency components usually contribute more to the visual quality. Thus, the lower frequency components and the DC coefficient should be encoded with high priority. However, the FGS codec is designed in such a way that the enhancement layer encodes the residuals bitplane by bitplane with regards to the amplitude of the residual only, rather than the importance of the frequency components. On the other hand, the base layer coding which codes the DC and lower-frequency components with a higher accuracy by using smaller quantization parameters will result in smaller residuals for the enhancement layer. Consequently, in contrast to the base, layer DCT coefficients characteristics, the important DC and lower frequency components may have smaller values in the enhancement la
Peng Shaomin
van der Schaar Minacia
Kelley Chris
Senfi Behrooz
Vodopia John
LandOfFree
Method of and system for activity-based frequency weighting... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method of and system for activity-based frequency weighting..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method of and system for activity-based frequency weighting... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3203139