A Novel Wipe Transition Detection Method Based on Multi-Feature

Wipe transition is an important type of gradual transition, but false and missed detection are very serious in most literatures due to noise and motion. In the proposed scheme, color sub-image and edge sub-image of each frame are decomposed by Db-4 wavelet transition. In detecting potential wipe transition, the color sub-image is divided into 8*8 pixel blocks, and Gaussian mode is applied to adjust the threshold dynamically, which is effective in reducing noise influence. To distinguish wipe transition from other gradual transition and motion, the edge sub-image is analyzed by Hough transform. More precisely, the wipe properties of completeness and independence in color sub-image are analyzed to locate the ranges of wipe transitions. The efficiency of the proposed method is extensively tested on news, football, MTV and other videos. The results show that 92% recall with 96% precision has been achieved, and the wipe transition frame ranges can be located precisely.