Scott Konishi

Learn More
—We formulate edge detection as statistical inference. This statistical edge detection is data driven, unlike standard methods for edge detection which are model based. For any set of edge detection filters (implementing local edge cues), we use presegmented images to learn the probability distributions of filter responses conditioned on whether they are(More)
We treat the problem of edge detection as one of statistical inference. Local edge cues, implemented by filters, provide information about the likely positions of edges which can be used as input to higher-level models. Different edge cues can be evaluated by the statistical effectiveness of their corresponding filters evaluated on a dataset of 100(More)
This paper investigates the use of colour and texture cues for segmentation of images within two specified domains. The first is the Sowerby dataset, which contains one hundred colour photographs of country roads in England that have been interactively segmented and classified into six classes – edge, vegetation, air, road, building, and other. The second(More)
— We propose a statistical approach to combining edge cues at multiple scales using data driven probability distributions. These distributions are learnt on the Sowerby and South Florida datasets which include the ground truth positions of edges. We evaluate our results using Chernoff information and conditional entropy. Our results demonstrate the(More)
We describe a novel viewpoint-lighting ambiguity which we call the KGBR. This ambiguity assumes orthographic projection or an affine camera, and uses Lambertian re-flectance functions including cast/attached shadows and multiple light sources. A KGBR transform alters the geometry (by a three-dimensional affine transformation) and albedo properties of(More)
We address the visual ambiguities that arise in estimating object and scene structure from a set of images when the viewpoint and lighting are unknown. We obtain a novel viewpoint-lighting ambiguity called the KGBR that corresponds to a group of three-dimensional affine transformations on the object or scene geometry combined with transformations on the(More)
| Recent work 1] has demonstrated the GBR ambiguity which is inherent in the perception of depth from shading and shadow cues. The purpose of this paper is rstly to extend the GBR transform to allow for diierent viewpoints. We demonstrate how this can be done by using image warping. This enables us to generalize the GBR to deal with situations where objects(More)
— We propose a statistical approach to combining edge cues at multiple scales using data driven probability distributions. These distributions are learnt on the Sowerby and South Florida datasets which include the ground truth positions of edges. We evaluate our results using Chernoff information and conditional entropy. Our results demonstrate the(More)
—We show that generic viewpoint and lighting assumptions resolve standard visual ambiguities by biasing toward planar surfaces. Our model uses orthographic projection with a two-dimensional affine warp and Lambertian reflectance functions, including cast and attached shadows. We use uniform priors on nuisance variables such as viewpoint direction and the(More)