Statistical Characteristics of Stereoscopic Images for Image Coding

Publisher Website
Google Scholar
Abstract
Statistical characteristics of stereoscopic images and the possibility of stereoscopic image data compression utilizing the mutual correlation between right and left images are presented. First, the mutual (cross) correlation between right and left images and the autocorrelation of the left images are measured. Next, one image is divided into blocks of fixed size. Each block is shifted consecutively and then subtracted from the corresponding area of the other image to form a residual block at each displacement position. The block with the least residual among all translated blocks is determined. These least residual blocks are then assembled to form a least residual image. Finally, the statistical properties of residual images and block translation values are investigated. The results indicate that data compression is possible by shifting one image horizontally and subtracting it from the corresponding area of the other. This research allows the efficient image coding for stereoscopic images by utilizing the correlation between two images, and makes possible low cost, more realistic 3D visual communications.

This publication has 0 references indexed in Scilit: