Performance Analysis between Basic Block Matching and Dynamic Programming of Stereo Matching Algorithm

Abstract

One of the most important key steps of stereo vision algorithms is the disparity map implementation, where it generally utilized to decorrelate data and recover 3D scene framework of stereo image pairs. However, less accuracy of attaining the disparity map is one of the challenging problems on stereo vision approach. Thus, various methods of stereo matching algorithms have been developed and widely investigated for implementing the disparity map of stereo image pairs including the Dynamic Programming (DP) and the Basic Block Matching (BBM) methods. This paper mainly presents an evaluation between the Dynamic Programming (DP) and the Basic Block Matching (BBM) methods of stereo matching algorithms in term of disparity map accuracy, noise enhancement, and smoothness. Where the Basic Block Matching (BBM) is using the Sum of Absolute Difference (SAD) method in this research as a basic algorithm to determine the correspondence points between the target and reference images. In contrast, Dynamic Programming (DP) has been used as a global optimization approach. Besides, there will be a performance analysis including graphs results from both methods presented in this paper, which can show that both methods can be used on many stereo vision applications.