Texture Analysis of Compressed Images

We introduce a new method of image texture analysis that is based on techniques that compress images in the form of weighted finite automata (WFA). We show that the WFA representation of images contain information about an image that is useful in analysing local texture and define a simple texture analysis method based on WFA. This method is the first texture analysis method that uses weighted finite automata and demonstrates the potential of WFA for further developments in texture analysis and segmentation. We include experiments on sample images showing the results of our technique and discuss certain advantages that our method has over other types of methods.