An Optimization Based Approach for Maximizing the Information Content of Keypoints Detected on a Digital Image

The task of finding point correspondences between two images of the same scene is one of the most challenging problems for a wide variety of computer vision applications. With regard to this, matching precision difficulties are related to the information content provided by the feature descriptors that are extracted from a set of interest points previously detected within a digital image. To overcome such difficulties, methods and techniques which aim to increase the information content (distinctiveness) of such feature descriptors must be developed. In this paper, an optimization-based approach for maximizing the information content provided by a set of feature vectors computed by the feature description method known as Spider Local Image Feature (SLIF) is proposed. In order to test the feasibility of the proposed approach, several state-of-the-art swarm optimization algorithms, such as Particle Swarm Optimization (PSO), Artificial Bee Colony (ABC), Firefly Algorithm (FA) and Social Spider Optimization (SSO) were implemented. The proposed experimental setup results show that the problem of maximizing the distinctiveness of a set of feature descriptors could be effectively modeled as an optimization problem, and as such, be solved by implementing popular optimization techniques.