Enhanced image coverage using evolutionary keypoint selection

Abstract

ABSTRACTCoverage of image features plays an important role in many vision algorithms such as homography estimation since their distribution affects the accuracy of the estimated homography. This paper presents an evolutionary algorithm, namely genetic algorithm, in order to select the optimal set of features yielding maximum coverage of the image. The coverage metric employed in the study is a robust method based on spatial statistics. A chromosome structure was designed to indicate whether the image features will be employed in the coverage computation or not. Genetic operators such as recombination or mutations were employed to search for different sets of features. The paper shows evaluation results with statistical tests on two datasets. Results indicate that the approach can find the set of features that generate higher coverage values and this finding was also confirmed by an accuracy test on the computed homography for the original set of features and the newly selected set. Results also demonstrate that the new set has similar performance in terms of the accuracy of the estimated homography with the original one. This approach has an additional benefit of using fewer number of features ultimately reducing the time required for descriptor calculation and matching.