A MULTISTAGE FAST MOTION ESTIMATION SCHEME FOR VIDEO COMPRESSION (TA-P3)

Author(s) : Jiancong Luo (The University of Texas at Arlington, USA) Ishfaq Ahmad (The University of Texas at Arlington, USA) Yu Sun (The University of Texas at Arlington, USA) Yongfang Liang (The University of Texas at Arlington, USA)

Abstract : A multistage motion estimation scheme is proposed. The scheme extracts video characteristics by first performing an online video analysis separately for foreground and background regions. Motion parameters are extracted and pass to the next stage. The next stage includes a mathematical model for the block distortion surface (BDS) that enables the algorithm to accordingly adjust its search technique. The search is performed on a precise search area adaptive to the statistical property of the motion vector prediction error. Due to its self-tuning property, not only does the proposed scheme adapt to scenes by yielding better visual quality but it also yields a lower computational complexity, compared with the other predictive motion estimation algorithms on standard benchmark sequences.

Menu