Interferent-detection-based video optimization method

Hu Shimin,Zhang Fanglue,Zhao Han,Wang Jue

2016-01-01

Abstract:The invention relates to an interferent-detection-based video optimization method. The method comprises: according to a time superpixel method, an interferent in a video is detected and location in the video is marked; a conversion matrix between each two adjacent frames in the video is obtained based on the collected video, thereby obtaining an original camera path formed by a plurality of conversion matrixes, a matrix conversion unit P is obtained and the matrix conversion P and each conversion matrix are multipled to obtain an optimized camera path, so that the camera can avoid the interferent and a main object can be guaranteed to be in a picture of the camera; the optimized camera path is divided into a plurality of segments and each segment completes one monotonous behavior; a key frame representing the path change in each segment is obtained and is combined with an adjacent key frame; and matching with each segment is realized by using a linear model, thereby obtaining a linear camera path of the collected video. According to the invention, the video path can be guaranteed to be simple and consistent; and the forming content of the video scene can be adjusted by removing the interferent, thereby improving the sense of beauty of the video.