Correlation-based stereoscopic disparity estimation with high-performance commodity hardware

Abstract

Stereoscopic depth reconstruction is the estimation of scene depth from at least two images of the scene taken at different viewpoints. The main computational problem is identification of matching pixels between the images. Correlation-based stereoscopic depth reconstruction algorithms are a group of algorithms developed to locate the matching pixels by comparing groups of pixels in both images. Although correlation-based algorithms can be used to compute disparity maps faster than most other algorithms, they cannot compute them at the speeds required for applications such as Image Guided Surgery which require disparity maps to be computed at rates in the region of 30 frames per second.