A Robust Approach for Camera Break Detection in Color Video Sequence

A convenient and eficient indexing system is as essential to managing video information as it is to managing tezt. Creating an indez requires partitioning the video into meaningful segments, assigning indez terms to individual segments and combining the indez terms to form a table of contents o r an inverted indez file. To date, video partitioning methods are still in the research stage. This paper introduces a new camera break detection algorithm, namely the Local Selective HSV Histogram Comparison. Color and spatial information are used in this new algorithm to complement the deficiency of the ezisting algorithms. The performance of four ezisting algorithms together with our proposed one are evaluated with image sequences of various scenarios, including camera movements, zooming, moving objects, deformable objects and video with degraded image quality. In addrtion, an adaptive thresholdtng technique is proposed to replace the general use of a single threshold for the camera break detection algorithms.