Low-Cost Real-Time Stereo Vision Hardware with Binary Confidence Metric and Disparity Refinement

Abstract

This paper presents a real-time stereo vision System-on-Chip (SoC) architecture for a depth-field generation processor as required in 3D TV applications. Dense Stereo Vision is a complex problem and uses many resources to achieve a high quality depth image. To reduce the needed resources, this architecture divides the problem into two parts: first a rough depth map is constructed using a segmentation based SAD window comparison, second a disparity refinement module identifies false matches and replaces them.