An efficient disparity estimation algorithm for stereoscopic image compression

Abstract

This paper presents an efficient block-based disparity estimation, called adaptive directional, limited search (ADLS) algorithm for stereoscopic image compression. This algorithm makes use of the homogenous direction property of the disparity and spatial correlation between adjacent blocks on epipolar lines. The major feature of this algorithm is that not only its computational load is decreased drastically for calculating the disparity vectors on account of its directional and limited computations compared to the full search (FS) but also, the objective performance is better than or close to the FS, due to homogenous disparity vectors. The simulation results demonstrate that the computational load can be reduced by about a factor 2/spl sim/10 compared to exhaustive FS approach and the PSNR is competitive with or slightly better than the FS.