METHOD AND APPARATUS FOR TRACKING THE MOTION OF IMAGE CONTENT IN A VIDEO FRAMES SEQUENCE USING SUB-PIXEL RESOLUTION MOTION ESTIMATION

摘要

The motion of image content such as a point of interest in a video frames sequence is tracked using sub-pixel resolution motion estimation in forward and/or backward direction. The motion of image content or a point of interest between a reference frame and a following frame is estimated, starting from an integer pixel position in the reference frame. If the motion vector has a position between integer pixel positions, the coordinates of that motion vector are replaced by an adjacent integer pixel position in the following or previous frame, and the error value between these coordinates and the replacement coordinates is stored. Thereafter the following or previous frame is defined as the current frame, the motion of the object is estimated between the replacement coordinates in the current frame and a following frame, and the stored error value is added in opposite direction to the corresponding motion vector, so as to get a corresponding precise motion vector. If the precise motion vector has a position between integer pixel positions in the following or previous frame, its coordinates are replaced by an adjacent integer pixel position in the following or previous frame, and the corresponding error value is stored, and so on.

主权项

1. A method for tracking the motion of an image content, such as a point of interest or an object or a group of pixels, in a video frames sequence using sub-pixel resolution motion estimation, comprising:
a) estimating the motion of an image content between a reference frame and a following or preceding frame, starting from an integer pixel position in said reference frame; b) if the end point of the corresponding motion vector has a position between integer pixel positions in said following or preceding frame, replacing the coordinates of said motion vector end point by the coordinates of an adjacent integer pixel position in said following or preceding frame, and storing the error value between said end point coordinates and said replacement coordinates; c) defining said following or preceding frame as a current frame, and estimating the motion of said image content between said replacement coordinates in said current frame and a following or preceding frame; d) correcting the following or preceding frame end point coordinates of the corresponding motion vector by using said stored error value in opposite direction, so as to get a corresponding precise motion vector.