Image Cluster Compression Using Partitioned Iterated Function Systems and Efficient Inter-image Similarity Features

When dealing with large scale image archive systems, efficient data compression is crucial for the economic storage of data. Currently, most image compression algorithms only work on a per-picture basis - however most image databases (both private and commercial) contain high redundancies between images, especially when a lot of images of the same objects, persons, locations, or made with the same camera, exist. In order to exploit those correlations, it's desirable to apply image compression not only to individual images, but also to groups of images, in order to gain better compression ratesby exploiting inter-image redundancies.This paper proposes to employ a multi-image fractal partitioned iterated function system (PIFS) for compressing image groups and exploiting correlations between images. In order to partition an image database into optimal groups to be compressed with this algorithm, a number of metrics are derived based on the normalized compression distance (NCD) of the PIFS algorithm. We compare a number of relational and hierarchical clustering algorithms based on the said metric.In particular, we show how a reasonable good approximation of optimal image clusters can be obtained by an approximation of the NCD and nCut clustering. While the results in this paper are primarily derived from PIFS, they can also be leveraged against other compression algorithms for image groups.