Research on computational vision focusses on the intersection between machine and human perception. Research concentrates on edge detection in human and machine, leading to a new theoretical account of optimal edge detection, which explains human blur detection better than all other competing models. Members of the group have also developed models of visual search which explain the variation in search speed, and the correlation between reaction times in search experiments.
Using classification images, we can estimate the "receptive field" humans use when detecting blur. The thick black line shows one such receptive field used to detect a blurred image which has the luminanceprofile shown in grey.
An optimal multiscale edge detector. Edges in the image on left are detected at a single scale (middle) or at multiple scales using an optimal detector (right). The multiscale detector picks up smooth or blurred edges that the single scale detector cannot. The optimal multiscale detector used here has been shown to account very well for human blur detection performance.