Abstract
In previous research it has been shown (Phillips and Todd VSS '03) that an integration of two-dimensional spatial frequency and coherent structure has a direct effect on subjects' ability to discriminate via texture in a matching task. What cues does this texture information provide when mapped to a three-dimensional heightfield? As well, how does performance differ when compared to their 2-D equivalent? We suggest that a change in threshold is expected due to the different information available in 3-D mappings. In a series of experiments, fields of uniformly distributed 3-D noise were presented in a 2AFC matching task. Spatial frequency was varied along with various sorts of structuring and de-structuring of the noise. Observers performed best at frequencies between 1 and 3 cycles per visual degree, consistent with this lab's previous findings in 2-D. Performance varies with the slant of presentation and the amount of structure available in the 3-D surface. Furthermore, there is an interaction between slant and structure such that coherence of structure is impeded by information that is introduced by 3-D viewing at higher amounts of slant. The simple nature of our 3-D heightfield mapping of the 2-D textures is that there will be approximately twice the ‘stuff’ (shading differences) used to denote the exact same number of ‘things’(coherent objects). These results beg the question — Just what is so-called texture, anyway? The phenomena of texture can arise from many sources including but not limited to pigmentation variations, geometric variations, illumination variations, and even motion. What is the interaction of these sources in the creation of phenomenal texture? Our results show that it is not simply ‘feature’ information but rather a more rich interaction of 2- and 3-D properties.