| With the continuous upgrade of video communication products and requirements, video coding technique is becoming an important bottleneck to the rapid development of multimedia communication. The current shortage media and network bandwidth condition have brought forward new challenge to video compression coding. Therefore, how to compress and coding to video data effectively is most important, which has already attracted considerable attentions recently. The wavelet analysis technology offers a self-adaptive localization analysis method for a given signal in both temporal domain and frequency domain, it can catch effectively no steady signal of video image and attain higher compression ratio. In addition, multi-resolution motion estimation and motion compensation can be done easily in the wavelet domain, and compare to traditional motion estimation in spatial domain, it can save more searching time. Mathematical morphology has unique predominance in image edge detection and video object segmentation, which can be computed simply and realized easily on the hardware, so it can be used with wavelet transform to enhance coding efficiency.In the dissertation, the characteristic of video image transmission on the very low bit rate is taken into count and the idea of video image segmentation is adopted, and then wavelet domain multi-resolution motion estimation on object and background is operated respectively. Based on research deeply on video object extract based on mathematical morphology and wavelet domain motion estimation algorithm, this paper proposes a new advanced algorithm. There are three main aspects as follow:Firstly, a new algorithm of video object extract based on watershed segmentation and MRF(Markov Random Field) classification is proposed. We adopt the idea of segmentation and extract to video object and make use of watershed segmentation method from morphology. Firstly, video image has been segmented into different regions and object and background have been separated by genetic descend method based on MRF. Secondly, we make use of erosion and dilation method to extract contour of video object and then attain video object.Secondly, based on video object segmentation result above, a new SPIHT (Set Partitioning In Hierarchical Trees) algorithm based on video object region, that is VOP-SPIHT, is proposed. We decompose video object in wavelet domain and make a set partition of spatial trees to wavelet coefficients, so as to improve compression ratio and transmission efficiency through scaleable encoding.And then the motion estimation algorithm is improved and a new motion estimation algorithm based on wavelet matching error characteristic in wavelet domain is proposed. The algorithm is based on PDE (Partial Distortion Elimination) and makes use of wavelet matching error characteristic and multi-resolution to realize motion estimation in the wavelet domain, as a result it improved the speed of searching best match.Finally lots of experiments with the proposed algorithm are operated and good results are attained. Then we make a conclusion to the article and discuss the further research of wavelet and mathematical morphology in the video coding. Taking mathematical morphology method directly into wavelet coefficients encoding may be another effective way. |
PDF Full Text Request