In the field of predictive video coding and format conversion, there is an increasing attention towards estimation of the true inter-frame motion. The restoration of motion vector field computed by 3-D RS is addressed...In the field of predictive video coding and format conversion, there is an increasing attention towards estimation of the true inter-frame motion. The restoration of motion vector field computed by 3-D RS is addressed and a propagating adaptive-weighted vector median (PAWVM) post-filter is proposed. This approach decomposes blocks to make a better estimation on object borders and propagates good vectors in the scanning direction. Furthermore, a hard-thresholding method is introduced into calculating vector weights to improve the propagating. By exploiting both the spatial correlation of the vector field and the matching error of candidate vectors, PAWVM makes a good balance between the smoothness of vector field and the prediction error, and the output vector field is more valid to reflect the true motion.展开更多
A globally optimal solution to vector quantization (VQ) index assignment on noisy channel, the evolutionary algorithm based index assignment algorithm (EAIAA), is presented. The algorithm yields a significant reductio...A globally optimal solution to vector quantization (VQ) index assignment on noisy channel, the evolutionary algorithm based index assignment algorithm (EAIAA), is presented. The algorithm yields a significant reduction in average distortion due to channel errors, over conventional arbitrary index assignment, as confirmed by experimental results over the memoryless binary symmetric channel (BSC) for any bit error.展开更多
This paper presents a single-pass bit allocation and rate control method for H.264. We first present an R-Q model which can estimate the frame codingcomplexity before macroblock mode selection. Based on our R-Q model,...This paper presents a single-pass bit allocation and rate control method for H.264. We first present an R-Q model which can estimate the frame codingcomplexity before macroblock mode selection. Based on our R-Q model, a single-pass slice layer rate control scheme is developed, including bit allocation of the frame layer and the slice layer. At last, to enhance our rate control scheme with greater accuracy, we propose a fine rate control method based on mode selection by optimizing the Lagrange multiplier. The experiments show that compared to JVT-E069 and JVT-G012 proposed to JVT (Joint Video Coding Expert Group), our methodcan control the bit rate more accurately and obtain a higher PSNR. When the three schemes applied to the sequence with scene changes the PSNR of this paper is improved by 0.2～0.4dB.展开更多
In recent years, more and more applications of video communication over the Internet have been extended, so the demand for reliable transmission of compressed video in a packet loss environment is ever increasing. Rat...In recent years, more and more applications of video communication over the Internet have been extended, so the demand for reliable transmission of compressed video in a packet loss environment is ever increasing. Rate-Distortion optimized mode selection is a fundamental problem of video communication over packet-switched networks, but the classical R-D method only considers quantization distortion in the source and hence it cannot achieve global optimality. Here we introduce a new global R-D optimal Macro-Block coding mode decision scheme for the new H.26L video compression standard. Based on the Internet packet loss model of Bernoulli and Gilbert, this R-D mode decision approach can result in better error robustness than classical method. Furthermore, our experimental results also demonstrate its superior adaptive error resilience and feasibility.展开更多
文摘In the field of predictive video coding and format conversion, there is an increasing attention towards estimation of the true inter-frame motion. The restoration of motion vector field computed by 3-D RS is addressed and a propagating adaptive-weighted vector median (PAWVM) post-filter is proposed. This approach decomposes blocks to make a better estimation on object borders and propagates good vectors in the scanning direction. Furthermore, a hard-thresholding method is introduced into calculating vector weights to improve the propagating. By exploiting both the spatial correlation of the vector field and the matching error of candidate vectors, PAWVM makes a good balance between the smoothness of vector field and the prediction error, and the output vector field is more valid to reflect the true motion.
文摘A globally optimal solution to vector quantization (VQ) index assignment on noisy channel, the evolutionary algorithm based index assignment algorithm (EAIAA), is presented. The algorithm yields a significant reduction in average distortion due to channel errors, over conventional arbitrary index assignment, as confirmed by experimental results over the memoryless binary symmetric channel (BSC) for any bit error.
文摘This paper presents a single-pass bit allocation and rate control method for H.264. We first present an R-Q model which can estimate the frame codingcomplexity before macroblock mode selection. Based on our R-Q model, a single-pass slice layer rate control scheme is developed, including bit allocation of the frame layer and the slice layer. At last, to enhance our rate control scheme with greater accuracy, we propose a fine rate control method based on mode selection by optimizing the Lagrange multiplier. The experiments show that compared to JVT-E069 and JVT-G012 proposed to JVT (Joint Video Coding Expert Group), our methodcan control the bit rate more accurately and obtain a higher PSNR. When the three schemes applied to the sequence with scene changes the PSNR of this paper is improved by 0.2～0.4dB.
文摘In recent years, more and more applications of video communication over the Internet have been extended, so the demand for reliable transmission of compressed video in a packet loss environment is ever increasing. Rate-Distortion optimized mode selection is a fundamental problem of video communication over packet-switched networks, but the classical R-D method only considers quantization distortion in the source and hence it cannot achieve global optimality. Here we introduce a new global R-D optimal Macro-Block coding mode decision scheme for the new H.26L video compression standard. Based on the Internet packet loss model of Bernoulli and Gilbert, this R-D mode decision approach can result in better error robustness than classical method. Furthermore, our experimental results also demonstrate its superior adaptive error resilience and feasibility.
