A comparison of different evolutive niching strategies for identifying a set of selfsimilar contractions for the IFS inverse problem

Abstract

The key problem in fractal image compression is that of obtaining the IFS code (a set of linear transformations)which approximates a given image with a certain prescribed accuracy (inverse IFS problem).In this paper,we analyze and compare the performance of sharing and crowding niching techniques for identifying optimal selfsimilar transformations likely to represent a selfsimilar area within the image. The best results are found using the deterministic crowding method.We also present an nteractive Matlab program implementing the algorithms described in the paper.The key problem in fractal image compression is that of obtaining the IFS code (a set of linear transformations)which approximates a given image with a certain prescribed accuracy (inverse IFS problem).In this paper,we analyze and compare the performance of sharing and crowding niching techniques for identifying optimal selfsimilar transformations likely to represent a selfsimilar area within the image. The best results are found using the deterministic crowding method.We also present an nteractive Matlab program implementing the algorithms described in the paper.

Keywords

• evolutive algorithms
• fractals
• iterated function systems
• image compression