The Evolution of Modern Texture Processing
This paper studies the evolution of image texture processing techniques over the last 20 years. Although texture is a fundamental attribute of images that has been shown to play an important role in human visual perception, the quantification and characterization of texture is difficult. Early texture processing techniques described texture deterministically or statistically in terms of repeated gray level patterns and the structure of the spatial placement of those 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 understanding of mammalian biological visual function that occured in the fields of psychophysics and physiology, and furthermore that the most successful modern methods is examined, and several of the most successful new techniques such as the multidimensional Teager-Kaiser operator and AM-FM modeling techniques are described in some detail. The use of computed dominant modulations to perform effective texture segmentation is demonstrated for the first time.
HAVLICEK, Joseph P. (1997) "The Evolution of Modern Texture Processing," Turkish Journal of Electrical Engineering and Computer Sciences: Vol. 5: No. 1, Article 1. Available at: https://journals.tubitak.gov.tr/elektrik/vol5/iss1/1