A study on graph-based moving object detection and tracking in an H.264/SVC bitstream domain for video surveillance

Abstract

This dissertation presents a spatio-temporal graph based method of detecting and tracking moving ob-jects by treating the encoded blocks with non-zero motion vectors and/or non-zero residues as potential parts of objects in H.264/AVC bitstreams and H.264/SVC bitstreams. In the H.264/AVC scheme, a spatio-temporal graph is constructed by first clustering the encoded blocks of potential object parts into block groups, each of which is defined as an attributed subgraph where the attributes of the vertices represent the positions, motion vectors and residues of the blocks. In order to remove false-positive blocks and to track the real objects, tem-poral connections between subgraphs in two consecutive frames are constructed and the similarities between subgraphs are computed, which constitutes a spatio-temporal graph. For the H.264/SVC scheme, an efficient detection and reliable tracking of real moving objects are first performed in the spatial base layer of H.264/SVC based on a spatio-temporal graph which is constructed from the block partitions with non-zero motion vectors and/or non-zero residue information. The spatio-temporal graph is utilized in reliably maintaining the real mov-ing objects of being detected and tracked by removing false detected objects via graph pruning and graph projection. Graph matching is then performed to precisely identify the real moving objects over time even under occlusion. Finally, region of interests (ROIs) of the detected objects are constructed as rectangles that encapsulate the block groups that represent the objects. To approximate the real shape and location of the objects, an ROI refinement process is performed. ROI refinement is also performed to determine the identification of two occluded objects during occlusion, which can be determined from the attributes of their vertices. For low-complex but accurate detection and reliable tracking of moving objects in spatial enhancement layer of H.264/SVC, inter-layer graph mappin...