Sparsity and Active Bases

Ying Nian Wu
University of California, Los Angeles (UCLA)

Sparse coding is a fundamental principle in image representation. In this lecture, I will explain Olshausen and Field method for learning sparse code from natural image patches. I will also explain related algorithms such as K-SVD. I will then describe a class of models called active basis models for representing image patterns of natural scenes. Finally I will describe a compositional sparse coding scheme for natural image data.

Presentation (PowerPoint File)

Back to Graduate Summer School: Computer Vision