Contrast enhancement influences the detection of gradient based local invariant features and the matching of their descriptors

Abstract

Contrast enhancement (CE) plays an important role in digital photography, medical imaging or scientific visualization, compensating for deficient dynamic range aspects. Our experiments show that CE via histogram modification influences the detection of gradient based local invariant features (LIF) and the matching of their descriptors. We bring evidence that the number of keypoints that can be automatically extracted by gradient based detectors increases with CE, and that matching gradient based keypoint descriptors extracted from image sets processed by CE is negatively affected in terms of Precision–Recall. We observed the effects of several classical and state-of-the-art CE methods on two widely used LIF detection/description techniques: Scale Invariant Feature Transform (SIFT) and Speeded-Up Robust Features (SURF).