Analysis and Evaluation of Keypoint Descriptors for Image Matching

Feature keypoint descriptors have become indispensable tools and have been widely utilized in a large number of computer vision applications. Many descriptors have been proposed in the literature to describe regions of interest around each keypoint and each claims distinctiveness and robustness against certain types of image distortions. Among these are the conventional floating-point descriptors and their binary competitors that require less storage capacity and perform at a fraction of the matching times compared with the floating-point descriptors. This chapter gives a brief description to the most frequently used keypoint descriptors from each category. Also, it provides a general framework to analyze and evaluate the performance of these feature keypoint descriptors, particularly when they are used for image matching under various imaging distortions such as blur, scale and illumination changes, and image rotations. Moreover, it presents a detailed explanation and analysis of the experimental results and findings where several important observations are derived from the conducted experiments.