Physics-based colour models in computer vision, colour science, graphics and multimedia

点击翻译按钮获取中文摘要

In my research, I incorporate detailed models based on the physics of interaction between light and surfaces to (a) understand images and video and hence the underlying lights and surfaces that created them; and (b) use these same effects in reverse to create realistic scenes in graphics.  I have developed a method  for generating (Computer Graphics) and understanding (Computer Vision) images using {\it whole spectra}, rather than just the usual Red, Green, and Blue of ordinary images. At the end of the day, in fact our visual equipment uses R, G, and B, but on the way, light is transmitted as a whole spectrum, including some red, green, etc., as well as every colour in between. So far, I have created a medical imaging application, for seeing inside volumes, that transforms magnetic MRI data into computer graphics. Since whole spectra are now available, some of the organs, bone, skin etc. can take on a unified colour (or no colour) while other organs are highlighted, and this can happen in real-time, while we use a slider.  I propose using the method to tackle difficult physics-based imagery such as the beautiful effects involved in insect iridescence, diamond facets, and the like.In Computer Vision, I have been looking at ways to {\it remove shadows} from images. This is very useful for such tasks as having a robot grasp an object without trying to grasp its shadow, trackingpeople in a security situation, etc. But it can also be used as simply a touch-up tool for images -- removing dark shadows on a face in a digital photo, say.  Shadows are useful in themselves, not just to be removed: they provide a profound cue about shape and lighting. I set out approaches in this proposal that help in the difficult issue of how to "re-light" images, by changing the shadows.