Using the finding that patches of roughly uniform 3D texture reveal the local orientation of the illumination we were able to map the global structure of the illuminance flow. We found that it is a robust indicator of the light field and an important entity for many subsequent inferences from the image such as shape from shading. Furthermore, through psychophysical experiments we found that human observers are able to estimate the local illumination orientation from 3D texture. The results of these experiments were in many regards similar to estimates on the basis of our computer vision algorithm, using the squared gradient or Hessian. Our latest findings seem to point in the direction that the squared gradient might be the effective stimulus for estimating local illumination orientation from 3D textures on globally flat surfaces by human observers. In this project we will extend these studies to 3D solid rough objects, with systematical variations of the shape, surface roughness and light field. Concrete aims are to 1. test shape and light field perception (psychophysics), and 2. compare the results of 1. with analyses of illuminance flow patterns (computer vision), for those 3D objects.

(This project covers part of Work Packages 1 and 5 of VISIONTRAIN project; goals are to investigate formal theories and test whether human vision models may or may not function along these theories for monocular visual cues and their interactions -WP1- and to provide a framework to compare formal mathematical model with psychophysical data -WP5-)