Feature Based Obstacle Detection Using Phase-Based Correspondence Matching and Image Declivity

This paper proposes a 2D feature based obstacle detection and calculating depth map by the disparity using phase-based correspondence matching and Image declivity. The phase information obtained from cross phase spectrum using 2D DFT (Discrete Fourier Transform) of images contains important information of image point similarity and dissimilarity. Level by level block matching using phase only correlation i.e. phase based correspondence search is used to find the correspondence point between input image and registered image using image declivity. Image registration is automatic, very fast, self adaptive segmentation technique as compare to other segmentation technique. I proposed reference point selection method based on image declivity which used for block matching and numbers of references point generated by image declivity is controlled according to application. The phase-based image matching is successfully applied to registered image for obstacle detection tasks for computer vision applications.