A Spatiotemporal Unsupervised Segmentation Algorithm for Video Object Plane Generation

• Main Authors: Ming-chieh Chan, 詹明傑
• Other Authors: E-Liang Chen
• Format: Others
• Language: en_US
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/87780634209318444012

碩士 === 立德管理學院 === 應用資訊研究所 === 94 === In this thesis, an efficient moving object segmentation algorithm suitable for implementing on a chip is proposed. The basic idea is to use simultaneously a novel spatial segmentation algorithm and an effective temporal algorithm based on accumulative information technique to produce the video objects. First, the spatial and the temporal segmentation algorithms are used simultaneously to produce the edges and the region of a moving object, respectively. Then, we take use of the edges with the region of the moving object to obtain the object. It is seen that the number of the time-consuming motion estimation computation can be reduced dramatically. Hence, the whole processing speed is significantly accelerated to meet the hardware implementation requirement. Finally, a post-processing step is applied on the object to smooth the object boundary. Good performance of this algorithm is demonstrated by the experimental results.