Comparing Images Based on Histograms of Local Interest Points

摘要

One of the key unresolved issues of image processing is the lack of methods for searching images similar to the reference image. This paper focuses on objects that there are in images and presents a method to compare the objects and search for images that contain objects belonging to the same classes. Taking advantage of the fact that local keypoints of images constitute a very good basis for further processing images, we use them for objects comparison. More precisely, the comparison of images is based on histograms, that are generated on the basis of the key-points of objects contained in images. We present results of experiments which have been conducted for various classes of objects and histograms generated using the proposed method.