Disparity estimation window size

Abstract

Abstract. Disparity estimation plays a crucial role in many stereo image compression techniques. To reduce computational complexity most methods limit the estimation search area to a limited window. The performance of the disparity estimation depends on the choice of the limited search window. Most techniques use a predetermined value for the window size, which is not optimal over a wide range of images. We show how the choice of the window size affects the performance of the stereo image compression algorithm and propose a method to obtain a better search window size. Our simulation results indicate an improvement of up to 1.81 dB over rigid window size selection and with performance very close to the optimal selection.