Doing


For the last three years I have been working on the development of the PennEyes binocular camera platform (PennEyes - A Binocular Active Vision System (1.3M), Technical Report MS-CIS-95-37/GRASP LAB 396). I have been using PennEyes to study issues in visual servoing and control such as redundant, three-dimensional tracking and image quality of moving objects captured from a moving platform.

Another project I am working on is the application of active vision to telepresence. We (in collaboration with Hany Farid) have developed a technique to acquire arbitrary views of a remote scene using an array of active cameras (Active Vision and Virtual Reality (3.3M), in: Exploratory Vision: The Active Eye, Michael S. Landy, Laurence T. Maloney and Misha Pavel (Eds.), Springer-Verlag, New York, 1995).

Using the electronic aperture facility available on some electronic cameras, I have developed a procedure to combine images taken with different temporal integration times into a single floating point representation. The Extended Intensity Range Images (1.1M) allow for the representation of the range of intensities found in real world images within a single image. Lambertian reflectance within deep shadow as well as intense specularity can be represented with equipoise.

I am currently working with the Laboratory for the Study of the Brain in Sleep in the Veterinary School applying signal processing techniques to the estimation of biological waveforms in the presence of overlap and noise. I have created a Matlab interface, called PoGO (abstract of a version presented at the World Federation of Sleep Conference, Nassau, Bahamas in September, 1995), for the analysis of pontogeniculooccipital waves recorded from the lateral geniculate nucleus of cats.

I have developed a computational model of early visual processing that extends the vector magnitude formulation of probability summation at detection to include discrimination. The model was initially applied to measures of spatial acuity and reconciled the sensitivity of observers to grating and resolution targets with that obtained with hyperacuity stimuli. The model is also applicable to the assessment of the spatial distribution of perceived differences in suprathreshold stimuli.


madden@grip.cis.upenn.edu / June 22, 1996