Real-time Stereo Rectification using Compressed Look-up Table with Variable Breakpoint Indexing

Abstract

Real-time rectification of stereo cameras is critical for stereoscopy in 3D vision and display. FPGA-based rectification has been suffering from the usage of large internal memory inside of the FPGA. The paper proposes a real-time stereo rectification approach using variable breakpoint indexing for enhanced compressed look-up table algorithm (E-CLUTR) in order to reduce the size of memory inside FPGA and to get a wider field-of-view (FoV) using high-distortion lenses. The proposed approach was successfully simulated using Altera ModelSim 10.4b. It is shown that the memory usage is reduced by 5865% compared with previous approaches and it supports real-time rectification of stereo 1920x1080 high-definition video data in 50 fps for a high-distortion lens with 110 degrees FoV.