Learning sparse multiscale image representations
Phil A. Sallee, Bruno A. Olshausen · 2003
We describe a method for learning sparse multiscale image repre-sentations using a sparse prior distribution over the basis function coecients. The prior consists of a mixture of a Gaussian and a Dirac delta function, and thus encourages coecients to have exact zero values. Coecients for an image are computed by sampling from the resulting posterior distribution with a Gibbs sampler. The learned basis is similar to the Steerable Pyramid basis, and yields slightly higher SNR for the same number of active coecients. De-noising using the learned image model is demonstrated for some standard test images, with results that compare favorably with other denoising methods. 1