Image Analysis in a Parameter-Free Setting

Authors: Yu Zhy and Thomas Zeugmann.

Source: “Information Sciences and Systems 2015, 30th International Symposium on Computer and Information Sciences (ISCIS 2015),”
(Omer H. Abdelrahman, Erol Gelenbe, Gokce Gorbil, and Ricardo Lent, Eds.), Lecture Notes in Electrical Engineering 363, pp.~285--294, Springer International Publishing 2015.

Abstract. The paper proposes a new method to approximate the normalized information distance by a compression method that is particularly suited for image data. The new method is based on a video compressor. The new method is used to compute the distance matrix of all the images in the data sets considered. Moreover, the hierarchical clustering method from the R package is used to cluster the distance matrix obtained. Two different datasets are considered to demonstrate the usefulness of our new image analysis method. The results are very promising and show that one can obtain a very good clustering of the image data.