In this paper, we propose a novel selective coding scheme based on layered structure. The scheme has 4 functional blocks that are target extraction, first layer coder for background region (FLC), second layer coder fo...
详细信息
ISBN:
(纸本)0780332598
In this paper, we propose a novel selective coding scheme based on layered structure. The scheme has 4 functional blocks that are target extraction, first layer coder for background region (FLC), second layer coder for target region (SLC) and buffer controller. For the target extraction process, we use motion vector grouping and change detection techniques, and increase the extraction accuracy of the target region by using a global motion compensation. The FLC and SLC are coders based on H.261 structure and generatelow resolution background and high resolution target image, respectively. Simulation results show that the average PSNRs are obtained about 45.05 dB and 26.02 dB for target and background region respectively. Also the average number of bits generated in a frame is about 4780 bits: 2.16 bits and 41.84 bits per block are assigned to background and target region, respectively.
Various neural network algorithms are introduced to implement vector quantization for image compression. These include competitive learning vector quantization, self-organizing feature map, frequency sensitive learnin...
详细信息
Various neural network algorithms are introduced to implement vector quantization for image compression. These include competitive learning vector quantization, self-organizing feature map, frequency sensitive learning, LBG neural network, and general learning vector quantization. A performance assessment of the above five typical neural networks is presented based on experimental results. A computation of the architecture design for a batch mode GLVQ algorithm is also presented, showing the best potential for further development.
The problem of designing algorithms without reference to future hardware implications is illustrated with respect to the implementation of H.261 and MPEG in VLSI systems. The Wavelet Transform is shown to be better su...
详细信息
The problem of designing algorithms without reference to future hardware implications is illustrated with respect to the implementation of H.261 and MPEG in VLSI systems. The Wavelet Transform is shown to be better suited to hardware implementation than MPEG and H.261 by an order of magnitude and achieves higher compression ratios for a given quality. A new mapping on to a VLSI architecture of the Wavelet Transform is described. A single device can process images at a frame rate of over 30 QCIF frames per second (8 bit pixels) and the compressed bit stream on average 33 kbits per second at an image quality of 32 dB (PSNR).
The framework of a novel algorithm for the detection, classification and identification of image subregions is introduced. It is shown that an efficient technique can be implemented for the representation of interfram...
详细信息
The framework of a novel algorithm for the detection, classification and identification of image subregions is introduced. It is shown that an efficient technique can be implemented for the representation of interframe motion in a video sequence. Classified subregion motion estimation produces a representation of motion that shows the displacement of regions, whilst employing classification to improve the efficiency of the prediction. The method can be used in motion vector coding in a simple, low-resolution video codec.
A new adaptive subband vector quantization (ASBVQ) codec for encoding video sequences at very lowbitrates in presented. Overlapped block motion compensation (OBMC) is employed to generate a prediction of the current...
详细信息
A new adaptive subband vector quantization (ASBVQ) codec for encoding video sequences at very lowbitrates in presented. Overlapped block motion compensation (OBMC) is employed to generate a prediction of the current frame from the previously decoded frame. The OBMC method generates a prediction error signal which can be compressed more efficiently when a subband based coder is used. A 2D discrete wavelet transform is applied to decompose the prediction error frame into 7 subbands. An adaptive subband VQ scheme is employed to quantize the wavelet coefficients and distribute the available bit budget among the subband vector quantizers.
A overview of the forthcoming ITU-T Recommendation H.263 is presented. In particular, the operation and impact on quality and bit-rate of the four H.263 Annexes (D, E, F and G) which specify optional enhanced modes of...
详细信息
A overview of the forthcoming ITU-T Recommendation H.263 is presented. In particular, the operation and impact on quality and bit-rate of the four H.263 Annexes (D, E, F and G) which specify optional enhanced modes of operation are discussed, with supporting results. The effectiveness of the various options is seen to depend upon the type of sequence and the frame rate, making it clear that for optimum performance a real codec must include means to make appropriate decisions about which options to invoke for maximum performance under given operating conditions. Finally, comparisons with H.261 indicate a significant performance gain for H.263, of the order of 2 to 3 dB.
In this paper a novel object oriented motion estimation algorithm is presented. The algorithm provides the means for highly efficient moving image encoding by fully exploiting the temporal redundancy among the objects...
详细信息
In this paper a novel object oriented motion estimation algorithm is presented. The algorithm provides the means for highly efficient moving image encoding by fully exploiting the temporal redundancy among the objects of successive frames. Two-dimensional segmentation is performed on a composite image synthesized from two consecutive frames. The object correspondence problem is removed implicitly by virtue of the fact that the generated composite segments correspond to successive versions of the same objects. Thus the scheme not only solves the problem of the correspondence between successive versions of the same object, but it also guarantees well matched segments in the presence of noise and varying illumination. Moreover, it preserves motion or deformation information. Progressive motion estimation is achieved within the segmentation process which adapts to the assumed translational or affine model. Motion compensated extrapolation is performed on uncovered background and overlapping regions of the predicted frame. Simulation results prove the efficiency of the predictive scheme even in the case that only motion and deformation parameters need to be transmitted.
A general-scene object-oriented image coder is presented which can encode colour TV signals at lowbit-rate independent of content, complexity, and temporal behaviour of the TV sequence. This is a novelty compared to ...
详细信息
A general-scene object-oriented image coder is presented which can encode colour TV signals at lowbit-rate independent of content, complexity, and temporal behaviour of the TV sequence. This is a novelty compared to object-oriented coders developed by other researchers which are specialized in video-phone scenes. For the realization of the coder's generality, stereoscopic and motion information has been used for segmentation and ordering of the objects. This coder is a new approach to the concept of object-oriented coding because it can encode without the use of pre-defined models and without a priori knowledge of the scene content. It is also unique in its type because it utilizes machine vision knowledge for object segmentation and manipulation. This work was done within the framework of RACE (Research on Advanced Communications in Europe) under contract R2045 `DISTIMA' (DIgital STereoscopic IMaging and Applications).
A new adaptive subband vector quantisation (ASBVQ) codec for the encoding of video sequences at very lowbitrates is presented. The block diagram of the complete system is shown. Overlapped block motion compensation ...
详细信息
A new adaptive subband vector quantisation (ASBVQ) codec for the encoding of video sequences at very lowbitrates is presented. The block diagram of the complete system is shown. Overlapped block motion compensation (OBMC) is employed to generate a prediction of the current frame from the previously decoded frame. The OBMC method generates a prediction error signal which can be compressed more efficiently when a subband based coder is used. A 2D discrete wavelet transform is applied to decompose the prediction error frame into 7 subbands. An adaptive subband VQ scheme is employed to quantize the wavelet coefficients and distribute the available bit budget among the subband vector quantizers. For applications where constant bitrate is required this is provided by the adaptive VQ algorithm, eliminating the need for a buffer.< >
A Hopfield neural network based block truncation coding (BTC) technique is presented in this paper. For this scheme, BTC is formulated as the minimization of a cost function in which the bit map distributions for the ...
详细信息
A Hopfield neural network based block truncation coding (BTC) technique is presented in this paper. For this scheme, BTC is formulated as the minimization of a cost function in which the bit map distributions for the blocks are explicitly included. It is explained that this cost function may also be interpreted as a measure of block detail. Based on the observation of the final value of the cost function found by the Hopfield network, a block may be classified as a high detail block or a low detail block, which are coded differently, giving a different compression ratio for each type. It is shown that using this new technique, compression ratios up to 7:1 with good reconstructed image quality can be achieved. Experimental results are presented to demonstrate the effectiveness of this new scheme.
暂无评论