Applying Histogram Contrast Analysis to Texture Perception

点击翻译按钮获取中文摘要

Many brain processes operate in parallel on sensory input to draw distinctions (boundaries, gradations, modulations, etc.) in the fabric of perceptual space/time. For instance, there are visual processes that operate on the spatiotemporal variations in light intensity to generate brightness distinctions. Other processes operate on variations in the spectral content of light over space and time to construct color distinctions. Still other processes draw distinctions based on spatiotemporal variations in textural properties (e.g., texture contrast, granularity and/or orientation). It is this last class of distinction-drawing processes, or channels, that these experiments will investigate. Histogram contrast analysis is a new branch of measurement theory whose purpose is to (i) isolate individual perceptual channels, and (ii) precisely characterize the computations those channels perform. No one knows exactly how many distinct channels are devoted to drawing texture-based perceptual distinctions, though there is general agreement that multiple channels are involved. Nor is there any data to indicate what specific computations are performed by any one of the putative channels used in texture perception. Heretofore, models of texture perception have relied on arbitrary, ad hoc guesses as to the nature of the channels used in drawing textural distinctions. These experiments will use histogram contrast analysis (i) to isolate individual channels used in drawing textural distinctions and (ii) to determine their computational characteristics. These experiments will thus make a start at the task of constructing a table of the computational "elements of texture perception" analogous to the periodic table of physical elements. Any theory of texture perception precise and comprehensive enough to predict performance probabilities on actual texture discrimination tasks must have such a table of elements as its cornerstone.