Sparse Intensity Histogram: Distinctive and Robust to the Space-distortion

Abstract

In this article, we propose an image descriptor which efficiently calculates the distance between images, saves to the disk compactly, and seeks a similar image related to a query robustly. The advantage of the proposed descriptor, sparse intensity histogram (SIH), is that it takes a robust approach to space distortion to the local descriptor, and that the speed of comparing are similar to the global descriptor because the SIH does not consider the spatial information, correspondence problem, to find the similar pairs of extracted descriptors between one and the other image. The experimental result shows that the proposed SIH has much better performance than the edge histogram descriptor in its accuracy.