Advances in Motion Detection and Tracking of Human as a Target

Abstract

Motion estimation of a target is the major area with higher computational complexity in video processing. It is the progression of discovery the motion patterns that describe the transformation from one frame to another in a sequence of video. Therefore, it is reasonable to carry out motion estimation only where movement is present. Image data in an image series remains mostly the same between frames along the target motion. To make use of the image statistics redundancy in image sequences, there is a need to guess motion. Motion estimation is valid for video compression improvement, stereo correspondence, object tracking and finding optical flow. Many precise methods have been proposed in the framework of one or more of these applications. Most motion estimation algorithms either operate directly in the image domain or finding the similar metric that measures how alike two pixels or two patches of pixels. In this paper, a review of a variety of motion estimation technique is presented.