Feature point detection in HDR images based on coefficient of variation

Feature point (FP) detection is a fundamental step of many computer vision tasks. However, FP detectors are usually designed for low dynamic range (LDR) images. In scenes with extreme light conditions, LDR images present saturated pixels, which degrade detection. On the other hand, high (HDR) no pixels but algorithms do not take advantage all information in such frequently relies on differential methods, work well HDR images, operation response bright areas overshadows dark areas. As an alternative to standard this study proposes detector based coefficient variation (CV) The CV adapts its deviation inside window, working both and proposed evaluated by measuring their repeatability rate (RR) uniformity. Our shows better performance when compared state-of-the-art detectors. uniformity metric, our surpasses algorithms. using worse than Harris SURF