SEMI-AUTOMATIC VIDEO OBJECT SEGMENTATION USING RECURSIVE SHORTEST SPANNING TREE AND BINARY PARTITION TREE

The objective of our work was to develop a fast and efficient tool for video content browsing and semantic video object extraction. The tool was developed using the Recursive Shortest Spanning Tree (RSST) algorithm and the Binary Partition Tree (BPT) technique. We first create an initial partition using the RSST algorithm which allows the user to specify the initial number of regions. We then progressively merge these regions to create the BPT thereby allowing the user to browse the content in a hierarchical manner. This merging step creates the binary tree with nearly double the userspecified number of homogenous regions. User interaction then allows grouping particular regions into objects. This user interaction is designed to allow object segmentation to be performed in a user-friendly manner. Any "interesting" regions can be marked in order to force them not to be further subdivided in the browsing process, which very importantly allows a small number of homogenous regions to be selected for an object. Other functionalities such as manually correcting the automatically generated results and multiple object segmentation, etc. are supported.