Tel-Aviv University - Computer Science Colloquium
Sunday, November 15, 14:15-15:15
at 14:00 Partial Differential Equations (PDE) are used extensively in image processing and computer vision. Many different PDE were suggested for image enhancement and segmentation in the context of scale-space and non-linear diffusion. I present the unifying and generalizing Beltrami framework who gives new answer to the question "what is an image?" and offers a new measure, the Polyakov action, on the space of images. It enables us to understand various algorithms as different limits of one master equation and to develop new algorithms which are state of the art in multi-channel (color, Texture and other attributes) image enhancement. A-priori knowledge about the class of images to be processed is easily and effectively incorporated in the Beltrami framework. Examples of grey-level, colored and textured images will be presented.
Based on joint works with R. Kimmel and R. Malladi and on joint work with Y. Y. Zeevi.