The Evolution of Modern Texture Processing

This paper studies the evolution of image texture processing techniques over the last years Although texture is a fundamental attribute of images that has been shown to play an important role in human visual perception the quanti cation and characterization of texture is di cult Early texture processing techniques described texture deterministically or statistically in terms of repeated gray level patterns and the structure of the spatial placement of these patterns Gray level cooccurrence matrices were among the most successful such methods Modern texture processing techniques tend to characterize texture in terms of spatio spectrally localized coherent amplitude frequency and phase modulations This paper argues that evolution of the modern methods from the early methods can be directly linked to advances in the understand ing of mammalian biological visual function that occurred in the elds of psychophysics and physiology and furthermore that the most successful modern methods have evolved to emulate biological vision systems Evolution of modern texture processing methods is examined and several of the most successful new techniques such as the multidi mensional Teager Kaiser operator and AM FM modeling techniques are described in some detail The use of computed dominant modulations to perform e ective texture segmentation is demonstrated for the rst time This research was supported in part by the Army Research O ce under contract DAAH and by the Air Force O ce of Scienti c Research Air Force Systems Command USAF under grant number F Author preprint Submitted to Elektrik Turkish Journal of Electrical Engineering and Computer Sciences