A WAVELET-BASED METRIC FOR VISUAL TEXTURE-DISCRIMINATION WITH APPLICATIONS IN EVOLUTIONARY ECOLOGY

Much work on natural and sexual selection is concerned with the conspi cuousness of visual patterns (textures) on animal and plant surfaces. Previous attempts by evolutionary biologists to quantify apparency of such textures have involved subjective estimates of conspicuousness or statistical analyses based on transect samples. We present a method b ased on wavelet analysis that avoids subjectivity and that uses more o f the information in image textures than transects do. Like the human visual system for texture discrimination, and probably like that of ot her vertebrates, this method is based on localized analysis of orienta tion and frequency components of the patterns composing visual texture s. As examples of the metric's utility, we present analyses of crypsis for tigers, zebras, and peppered moth morphs.