METHOD FOR DETECTING SPLITTING CAMERA MOTION IN COMPRESSED REGION OF COMPRESSED VIDEO

摘要

PURPOSE: A method for detecting and splitting a camera motion in a compressed region of compressed video is provided to partially decode compressed video so as to detect a camera motion finely splitted, thereby reducing the amount of calculation and attaining stabilized noise-resistant performance. CONSTITUTION: To automatically detect a camera motion from a video compressed by a compression method that compensates for motion in blocks, motion vectors in blocks are detected from the compressed video first. Then, a motion vector field of each frame forming a video sequence is constructed using the motion vectors. An affine motion model parameter that describes total motion is estimated from the motion vector field of each frame. The camera motion is detected by making the affine motion model parameter be a critical value(S104).