| Summary: | 博士 === 國立中央大學 === 電機工程學系 === 101 === To solve huge amount of interconnection between multimedia de- vices, wireless digit home systems, such as wireless HD, WHDI, and UWB, are built to meet the digit home requirement. How- ever, traditional wireless system focus on point-to-point high-speed transmission, while the distance and the data rate is severely limited. Therefore, a unified wireless digit home environment is pro- posed, named wireless high definition video interface-mobility, multiple access, and multimedia (WHDVI-m). In this thesis, an integrated scalable video processing system is proposed, inspired by the WHDVI-m system.
First, visual-quality particle based frame rate up conversion algorithm is proposed (named particle based frame rate up conversion) by combining the hierarchical MV field construction stage and PMTC. Estimated MVs from the hierarchical MV field construction stage are memorized and passed to PMTC to retrieve true MVs. Beside distance minimization procedure, joint evaluation of cost-function prevents untrue MVs from. According to the experiment results, proposed PFRUC outperforms existing works by 0.9 ̃3.9 dB, including bi-directional and unidirectional approach. The experiment results prove the necessity of unidirectional motion. On the other hand, PMTC, inherited from the bi-directional FRUC concept, is efficient in differentiating the true MVs from untrue ones.
Second, new cost evaluation is formulated in hybrid frame rate up conversion to replacement original cost evaluation in particle based motion trajectory calibration (the second stage of PFRUC). Repeated operation in particle based frame rate up conversion is removed by initial motion vector assignment. Subsequent motion vector mapping is designed for three purposes, including noise minimization, search range expansion, and computation complexity reduction. According to the experiment results, the visual quality is enhanced by 0.3 dB, compared with state of the art unidirectional motion estimation based frame rate up conversion, while those process higher visual quality compared with bi-directional based approaches. On the other hand, the computation complexity of hybrid frame rate up conversion is less than existing works by over 29% and 90% in processing CIF and full-HD videos.
Finally, a new lossless compression codec is proposed, named multiple- source scalable video compressor. Overall system contains the dis- crete wavelet transform, quality driven bitplane sequencer, and bi- nary arithmetic coder with some specific considerations. In binary arithmetic coder, M-coder is adopted to enhance the hardware realization performance, while the compression is slightly enhanced. Besides, two-pass quality driven bitplane sequencer is proposed based on the distribution of zeros and ones in significance layer. According to the experiment results, the computation time of pro- posed scalable image/video compressor is one third of openJPEG (JEPG 2000) while the average compression ratio of the proposed work is only 0.04 behind.
|
|---|
