Search

In this paper, we analyze the bit rate distribution of the Full Search and the fast search motion estimation algorithms in H.264. It shows both experimentally and theoretically that a well-designed fast block-matching algorithm not only speeds up motion estimation process, but also improves the rate-distortion performance. This is in contrast to the common belief that Full Search always yields optimal rate-distortion performance under the same test condition. (1) Index Terms--motion estimation, rate-distortion, Lagrange cost, H.264/AVC.

In this paper, we presents a PSNR-based frame complexity estimation to improve H.264/AVC rate control. Our scheme is based on adding PSNR-based complexity estimation to the existing mean absolute difference based (MAD-based) complexity measure to form combined frame complexity estimation. The aim is to allocate bits more accurately, especially for frames with scene changes and high motions. Bit allocation to each frame is not just computed by encoder buffer status but also by a combined frame complexity measure. The quadratic rate-quantization (R-Q) model is used to determine the quantization parameter (QP) for each frame and the computed QP is further adjusted in two special cases. We also propose a frame skipping decision scheme to improve the performance at high frame rates. Simulation results show that the H.264 coder, using our proposed scheme, achieves better visual performances with similar or smaller PSNR deviations when compared to that of the recent JM8.6 rate control. Index Terms--AVC, Bit allocation, H.264, Rate control, JVT, Video coding, PSNR

This paper presents a method for automatic optimization of FAPs (Facial Animation Parameters) based on a face video sequence. An FAP optimizer unit compares a rendered face (1) with the original face in the video sequence. The difference is iteratively minimized to produce an optimal set of FAPs. Unnecessary FAPs are eliminated and the search is partitioned to speed up the optimization process. Three different search techniques, Steepest Descent Method, Linear Search Method, and Cyclic Coordinates Method are applied to derive an optimum. The resulting FAPs are very accurate and the animation based on the FAPs is lifelike and truthful to the original video. Index Terms--MPEG-4, Facial Animation, FAP generation.

In Mpeg-2 HDTV (MP@HL) application, bandwidth and size of memory requirements are very large. These requirements increase bus bandwidth and hardware cost. We present a simple method that requires low complexity hardware for MPEG-2 decoder memory reduction. Our algorithm is based on Haar Wavelet Transformation and Haar basis with full-tree decomposition transformation. No additional multipliers are needed. Following the transformation, we use a variable-length coder to compress the data before storing them to the memory for future references. For memory accesses, we apply group of macroblock segment coding and hierarchical access method to reduce access time. According to our simulation results, our technique can at least cut down the memory requirement from 12 Mbytes to 7 Mbytes for 4:2:0 HDTV real-time application. The real performance is dependent on the bid rate and the original MPEG2 decoded video quality.

A video decoder with an efficient block-level-pipeline controller scheme for MPEG-2 MP@ML is presented in this paper. The architecture in most of the reported literature for MPEG-2 MP@ML video uses a 64-bit bus and a complex bus arbitration scheme to communicate with external DRAM, display, and incoming FIFO. Our design imposes a certain order in the DRAM access by various processing units instead of allowing any processing unit within the decoder to request bus access arbitrarily. This efficient DRAM accessing order allows us to reduce bus width from 64 bits to 32 bits, without significantly increasing embedded buffer sizes, and still meet the requirements for MPEG-2 MP@ML real-time decoding. The bus arbitration algorithm is also simple, allowing for a less complex controller design.

In this paper, we present an approach to increase the amount of VQ codebook patterns to improve image quality without a large overhead on the codebook memory and search time. Basically, our approach characterizes a pattern into a key value, which is used in codebook search; 'Reflection operation' was performed on codewords using the key values to get more patterns to improve image quality. Our basic VQ system is a 2-step VQ where VQ is used to compress prediction errors resulting from simple scalar quantization of pixel blocks. The codebook search time for our technique is O(logN+K) for each input vector, where N is the codebook size and K is the Voronoi's range. Experimental results are presented in the paper.