Enhancing image registration performance by incorporating distribution and spatial distance of local descriptors

摘要

A data dependency similarity measure called m(p)-dissimilarity has been recently proposed. Unlike l(p)-norm distance which is widely used in calculating the similarity between vectors, mp-dissimilarity takes into account the relative positions of the two vectors with respect to the rest of the data. This paper investigates the potential of mp-dissimilarity in matching local image descriptors. Moreover, three new matching strategies are proposed by considering both l(p)-norm distance and m(p)-dissimilarity. Our proposed matching strategies are extensively evaluated against l(p)-norm distance and m(p)-dissimilarity on a few benchmark datasets. Experimental results show that m(p)-dissimilarity is a promising alternative to l(p)-norm distance in matching local descriptors. The proposed matching strategies outperform both l(p)-norm distance and m(p)-dissimilarity in matching accuracy. One of our proposed matching strategies is comparable to l(p)-norm distance in terms of recall vs 1-precision. (c) 2018 Elsevier B.V. All rights reserved.