Fast Total Variation-Based Image Reconstruction

We apply a total variation minimization technique for two tasks in image processing: denoising, which is the reconstruction of an image from noisy observations of the image; and deblurring, which is the reconstruction of an image which is convolved with a smooth kernel function and then contaminated with error. Total Variation minimization yields a nonlinear elliptic partial diierential equation (PDE). In this paper we discuss numerical techniques for the eecient solution of this PDE.