Image texture analysis using weighted finite automata
Helmut Jürgensen, Mark Eramian · 2003
We investigate the viability of extracting local texture information from finite automata that encode images. Local Hausdorff dimension is considered as a possible measure of texture complexity and is found to be only somewhat representative of texture despite previous claims. A theoretical framework for analysis of images encoded by weighted finite automata is developed, implemented and tested. During this project, a new efficient algorithm for simulating weighted finite automata is introduced. We find that texture analysis of images from the finite automata that represent them is possible and the successful results of a first method for doing so are presented.