Survey on image rectification on FPGA

Abstract

Stereo vision is a technique for building a 3D description of a scene observed from two or more view points. It is most often used for many applications in robotics including 3D object recognition, localization as well as 3D navigation of mobile robots. Applications that are based on stereo vision exploit the correspondence between pixels between images at hand. However, this correspondence problem is one of the major issues in stereo vision, as it requires the search for matches in a 2D area but this problem can be reduced to 1D line search if images were rectified. Stereo image rectification is the process of transforming images to obtain image planes to be parallel, epipolar lines to be collinear and parallel to one of the image axes (usually the x-axis). Several systems have been successfully implemented to perform real time image rectification on FPGA. Most of these implementations use Look Up Table (LUT) approach on a calibrated stereo rig