Visual Perception and Coding of Texture

点击翻译按钮获取中文摘要

DESCRIPTION (provided by applicant): The experiments in this proposal share two basic themes. The first is that visual texture stimuli may be used to probe the initial visual coding of images and image sequences. The second theme is to relate all of our work to a basic 4-stage, multiple channel model, the "back-pocket model of texture segregation." Each channel consists of (1) an initial linear spatial filter, (2) a nonlinearity, and (3) a second linear filter. These channels feed (4) a combination rule or linking hypothesis that yields a response or prediction of performance in a given task. Our work attempts to confirm this model, characterize its mechanisms, analyze its capabilities, and relate it to the underlying physiology. This proposal has three major aims. The first aim concerns the coding and appearance of natural textures. We investigate whether human texture coding is matched to the statistics of natural textures and the degree to which observers can estimate physical characteristics of textures independent of viewing conditions. The second aim investigates how the back-pocket model channels are used for the identification of texture defined objects and includes further investigations of the structure and selectivity of back-pocket model channels. The third aim is to investigate the cortical implementation of texture or 2nd-order channels using fMRI to measure responses to and adaptation to texture modulations.