Perceptual texture space improves perceptual consistency of computational features

Abstract

Perceptual consistency is important in many computer vision applications. Unfortunately, except for color, computational features and similarity measurements for other visual features are not necessarily consistent with human's perception. This paper addresses three critical issues regarding perceptually consistent texture analysis: (1) development of perceptual texture space, (2) assessment of how consistent computational features are to human perception, and (3) mapping computational features to perceptual space. It demonstrates the construction of a reliable perceptual texture space, which can be used as a yardstick for assessing the perceptual consistency of computational features and similarity measurements. Moreover, it is found that commonly used computational texture features are not very consistent with human perception, and mapping them to the perceptual space improves their perceptual consistency.