Research On Moving Video Object Segmentation And Advanced Motion Estimation/Motion Compensation Algorithms

Video object segmentation can separate video frame into many video objects with vido analysis tools in temporal and spatial domain. It is a key pre-processing for content-based interactive video processing systems, such as content-based video compression, indexing, representation, and retrieval systems, and is a very important component in MPEG-4 and MPEG-7. At the same time, demand for enhanced compression efficiency from interactive (video telephony) and non-interactive applications (broadcast, streaming multimedia, stroage, video on demand) is still stringent. Improvement of motion estimation/motion compensation (ME/MC) techniques has been the major reason for coding efficiency improvements achieved by modern standards, when comparing them from generation to generation. So, in this dissertation, moving video object segmentation and advanced ME/MC algorithms are studied.Efficient, accurate and automatic segmentation of video object is a key technology in object-based video coding. An automatic and accurate moving object extraction algorithm is proposed to segment head-shoulder video sequence. First, an improved background registration technique is provided to construct a reliable background image. The moving object region is then detected by combining the background difference with the frame difference. Then, a new temporal filter and a post-processing step are applied on the detected change to improve the temporal coherence and spatial integrity. Finally, an improved watershed algorithm is proposed to locate the precise video object boundary, with significant reduction of computation compared to conventional watershed algorithm. Experimental results show that the proposed method can segment the head-shoulder video object with accurate object boundary and a processing speed of 11 QCIF frames per second can be achieved on a PC.In low bit-rate video surveillance application, the image quality at moving object region (region-of-interest, ROI) is usually poor. To cope with this problem, we propose a ROI video coding framework, which gives higher priority to the moving object region. A fast and automatic moving object extraction algorithm is adopted to dynamically define ROI. A quality-rate joint control method is applied toROI coding to enhance the quality at ROI. The ROI video coding framework is implemented on H.263+ encoder. Experimental results show that proposed method can significantly improve quality at ROI, while sacrificing the quality on background. So, the proposed ROI video cding strategy is very suitable for real-time low bit-rate video surveillance application.H.264/AVC is the latest standard for video coding, which provides gains in compression efficiency of up to 50% at the expenses of an increased implementation cost up to one order of magnitude for the encoder, compared to previous standard. In order to reduce the heavy computation burden of H.264 encoder, two fast and effective mode decision algorithms, based on local motion activity estimation, are presented. The first algorithm is based on frame difference and displaced-frame-difference analysis, whichs utilizes the spatial homogeneity of video object's textures, the motion activity information and the statistical characters on reference frame selection in motion estimation, to effectively skip unnecessary trials of coding modes and reference frames. Experimental results show that the proposed method achieves a reduction of 40-80% encoding time, with a negligible PSNR loss and bit-rate increase.In order to offer the H.264 encoder a trade-off between picture quality and the embedded available computational performance, a fast mode decision algorithm with adjustable computational-complexity is presented. According to the analysis of the characteristics of the coding-modes and statistical characteristics about the best mode, the proposed algorithm exploits motion activity and the quantized transform coefficients of the residue, obtained by the tested coding-modes, to effectively predict the coding efficiency of the remaining modes and skip the unnecessary modes. Thresholds used in this algorithm can be adjusted to regulate computational complexity. Experimental results show that the speed of the proposed method is about 2-4 times faster than the full-mode decision algorithm without any noticeable loss of image quality and coding efficiency. The proposed complexity scalability algorithm is helpful for mobile consumer devices using MPEG video coding.To reduce the computational complexity of variable block-size motion estimation in H.264 encoder, a fast algorithm is presented, which effectively utilizes the motion activity information. The initial search center location is predicted accurately by exploiting the features of the variable block-size motion estimationand the spatial-temporal correlation of motion vectors. In order to skip unnecessary search points, a search pattern selection strategy is provided, which can adapt to the local motion activity. Experimental results show that, compared to the H.264 reference software JM8.4, the proposed method reduces 65% of the search points on average with the same compression efficiency.