System and method for semantic video object segmentation

摘要

The segmenting and actively tracking of semantic video objects from a sequence of frames of digital video information involves receiving an initial frame of video image data and at least one succeeding frame of video image data from a sequence of frames of digital video information. From the first frame, at least one object boundary is defined by a user. The initial frame of video image data is then partitioned into a plurality of regions based on selected substantially homogenous video features. Tracking is then performed on the object boundary and plurality of regions which are projected to at least one succeeding frame of the video image data.