Accelerated Feature Descriptor Matching in Images Using Threshold Accepting

Abstract

The process of matching two visual descriptions of images is a major task in Computer Vision. This matching is generally done using Exhaustive search (Brute-Force) and Nearest Neighbor search which has been proved computationally expensive in some cases. This paper proposes a heuristic method to perform feature descriptor matching. The heuristic approach applied here works based on a combinatorial optimization algorithm called Threshold Accepting. The experiments performed suggest that the proposed algorithm can produce better results within a minimum number of iterations than existing algorithms.