Blind deconvolution (with previously estimated kernel)

Input image Reset to Original vs AR Sparse prior using estimated kernel (Fergus et al) and λ = 0.01 Set as Left Image

Richardson-Lucy IID Gaussian AR Gaussian IID Sparse AR Sparse λ=0.01 λ=0.001

These examples illustrate blind deconvolution using a blur kernel estimated from the input image as a preliminary step. Both images and the corresponding estimated kernels are from Fergus et al (2006), Removing camera shake from a single image. The remaining parameters have their usual interpretations.