Autoregressive Models for Image Coding

Recently, the image and video coding community has witnessed several proposals to improve coding efficiency by exploiting perceptual redundancy of texture. Most of these approaches are based on segmentation and non-parametric texture models popular in the computer graphics domain. Although not a generic model for everything we might call texture, the simple (and parametric) autoregressive model has some properties that make it appealing for coding purposes. This talk is about a different approach to the problem based on this model, superposition, and denoising. Viernes 18 de Febrero de 2011, 12:00 h Sala de Juntas, Instituto de Óptica, CSIC C/ Serrano, 121, Madrid Contacto: J. Portilla, Tfno: 91 561 6800 x942333 Denoising Model Estimation Conv. Coder