Video Object Tracking in Compressed Domain Using SIFT and STMRF

The compressed video object tracking are required for a following structure with both reasonable accuracy and complexity. The existing systems of video object tracking which doesn't provide good accuracy. Up-till now all systems assumes that video is already in the H.264/AVC format. All these types of errors are overcome in this video object tracking system. Here we are presenting video object tracking in compressed domain by using two algorithms Scale Invariant Feature Transform (SIFT) and Spatio-Temporal Markov Random Field (STMRF) algorithm. Here for tracking we use YUV video format but if input video is in avi format then we convert this video into YUV format. Scale Invariant Feature Transform (SIFT) perform object detection using frame difference method. SIFT tracking has steps such as scale space extrema, key point localization, orientation assignment, keypoint descriptor and match key point and so forth. STMRF utilizes just the movement vectors (MVs) and block coding modes from the compressed bit stream to perform tracking. it is start with, preprocessing the MVs are preprocessed through intra coded block motion estimation and motion compensation. Global Motion Estimation (GME) is utilized to take a short moments of camera and Global Motion Compensation (GMC) is utilized to remove Global motion which include intra-coded block processing. At every frame, the choice of whether a specific block belongs to object decided with the help of the ST-MRF, which is updated from frame to frame. STMRF gives preferred exactness over SIFT.