Estimating the minimum redundancy in stereo image pair

The raw data in binocular stereo image sequences is twice as that of monocular images, the large amount of information should be reduced. As a result there has been increasing attention given to image compression methods specialized to stereo pairs. Much of this work has concentrated on improving the disparity compensation process and codes the residual image similarly to a monocular image where one view is used to predict another, and the difference is coded. The residual image is usually composed primarily of strong vertical direction edge components surrounded by large areas of near zero intensity. The residual images have different characteristics, but they behave uniquely statistical regularity. This property is demonstrated experimentally in the paper. Two interested statistical variables are described, the one is the total number (N) of the pixels with near zero intensity in the residual image and other is the coordinate displacements (Δx, Δy) between the left and right image frames for get the residual image. Experimental results indicate that the curve between the parameters N and variables (Δx, Δy) may be fit by Gaussian function. The maximum of the variable Nm corresponding to the optimal displacements (Δxop, Δyop) may be estimated by the Gaussian approximation. An algorithm is further provided to quickly predict the minimal redundancy of the residual image and the corresponding displacement. It is shown how such characteristics may be of great benefit to quickly achieve the higher compression ratio.