Colloquia / Seminars / Industry Talks

For about the last 10 years, the digital camera market has been obsessed with the number of pixels. New cameras can have 10, 12, 20 or more Mega-pixels per image. Yet, the need for so many pixels is questionable: digital cameras now have around twice the number of colour receptors that we ourselves have and we see the world in perfect clarity. Moreover, from a signal processing point of view, too many pixels means that, even in relatively normal lighting conditions, individual pixels often capture just a handful of photons. As a consequence the resulting raw images measured on a camera CCD are often very noisy. While this noise can be removed, by averaging proximal pixels, doing so reduces the effective resolution of the camera. At the University of East Anglia, and now in a spin out company called Im-Sense Ltd, we have been developing technology for making better looking images. Our approach is not to focus on the number of pixels in they eye or indeed the eye's general physiology. Rather, we have developed a unique mathematical algorithm which delivers images that look like those we ourselves remember seeing: the processed images are more appealing and vivid (in fact, they appear to have higher definition). Our approach is entirely consistent with our own vision system where the eye records an image but the picture we see is the result of extensive cortical processing. Thus, we believe a major part of the image quality story will be the processing of images and not just the pixel count of the acquisition device or the resolution of the display. In this talk I will review some of the major processing approaches that have evolved (including Retinex theory and ICAM) and give my view of why these approaches are limited in their application. These limitations are the starting point of the UEA/Im-Sense approach. A high level outline of our algorithm will be presented along with experimental results.

Graham D. Finlayson gained a BSc degree in computer science (with honours) from the University of Strathclyde in 1989, and MSc and PhD degrees from Simon Fraser University, Vancouver, Canada, in 1992 and 1995. His PhD dissertation was warded a Dean's medal for academic excellence. From 1995 to 1997, he was a lecturer in the Department of Computer Science at the University of York in the United Kingdom. In 1997, he was appointed to a readership at the University of Derby, where he was a founding member of the Colour & Imaging Institute. In 1999, he left Derby to become Professor in the School of Computing Sciences at the University of East Anglia. Professor Finlayson is interested in how color can be used to solve problems in computer vision and allied disciplines such as image processing and digital photography. He has made many contributions toward solving the color constancy problem (removing color bias due to illumination from images) and his algorithms have been implemented in commercial cameras. His current research interests include dynamic range compression, the automated removal of shadows in images, photometric invariance, and the application of computational techniques to the understanding of human vision. Professor Finlayson is also the Chief Technical Officer of Im-Sense Ltd.