Reconstruction-Diffusion: An Improved Maximum Entropy Reconstruction Algorithm Based on the Robust Anisotropic Diffusion

Abstract

Maximum entropy (MENT) is a well-known image reconstruction algorithm. If only a small amount of acquisition data is available, this algorithm converges to a noisy and blurry image. We propose an improvement to this algorithm that consists on applying alternately the MENT reconstruction and the robust anisotropic diffusion (RAD). We have tested this idea for the reconstruction from full-angle parallel acquisition data, but the idea can be applied to any data acquisition scenario. The new technique has yielded surprisingly clear images with sharp edges even using extremely small amount of projection data.