PSO Authenticated Active Mesh Based Motion Estimation and Compensation in Wavelet Sub-bands

Authors

  • Vidyasagar K N, Bharathi S H

Abstract

Motion Estimation and Compensation in video processing can be achieved by a sub-pixel full search algorithm with accuracy up to 1/4th pixel integer point with a reduced coding rate of an encoder. Intended at this problem, a new method for motion estimation using Wavelet-based-Contourlet-Transform (WBCT) in sub-pixel level is proposed. To begin with the optimal low-frequency integer pixel points are calculated in sub-band of the video frame is, and then by considering the optimal integer pixel as a centre of the hunt, by using chaotic-Particle Swarm-Optimization (PSO) method. The primary motion-vector is found out analogous to higher frequency sub-band of the video frame. Finally, the low-frequency integer pixel points are authenticated as feature points by using improvised Lucas-Kanade optical flow method, then to avoid the aperture problem and chaotic nature of feature points within wavelet sub-bands. An optimized Active Mesh Motion Compensation is incorporated. The Proposed algorithm has experimented on video sequences taken from the standard database and the result reveal that the use of projected optimised Active Mesh algorithm for Motion Compensation in Wavelet sub-bands within the video frame shows a good improvement in PSNR performance. A good Video Structural Similarity Index Measure (VSSIM) can be achieved by the proposed algorithm for the reconstructed video frame when compared with conventional methods.

Published

2020-11-01

Issue

Section

Articles