KVD: Scale invariant keypoints by combining visual and depth data

Abstract

One of the first steps in numerous computer vision tasks is the extraction of keypoints in images. Despite the large number of works proposing image keypoint detectors, only a few methodologies are able to efficiently use both visual and geometrical information. In this work we introduce KVD (Keypoints from Visual and Depth Data), a novel keypoint detector which is scale invariant and combines intensity and geometrical data using a decision tree. We present results from several experiments showing that the detector created with our methodology outperforms state-of-the-art methods, both in repeatability scores for rotations, translations and scale changes, as well as in robustness to corrupted visual or geometric data. Additionally, as far as processing time is concerned, KVD yields the best time performance among the methods that also use depth and visual data.