Detecting scene changes on encoded video sequences.

摘要

Encoded video sequences are processed prior to complete decoding. Scene changes are readily detected using DCT coefficients in JPEG and MPEG encoded video sequences. Within each 8x8 block, the distribution of the DCT coefficients is used to classify the block as either type O (contains no high frequency components) or type 1 (contains high frequency components) The changes in the distributions of O's and 1's from one frame to next is captured using eigenvectors and used to represent scene changes. The frames in between two consecutive scene changes form a video shot. Video shots may be thought of as the building blocks of video sequences, and are used in browsing, database indexing, or any other operations that essentially form an abstraction of the video. To visualize each video shot, the contents are abstracted in a meaningful manner such that it is the representative of the contents of the shot; this is achieved using representative frames or Rframes.