Robust Algorithm for Denoising of Photon-Limited Dual-Energy Cone Beam CT Projections

Dual-Energy CT offers significant advantages over traditionalCT imaging because it offers energy-based awareness of theimage content and facilitates material discrimination in theprojection domain. The Dual-Energy CT concept has intrinsicredundancy that can be used for improving image quality, byjointly exploiting the high- and low-energy projections. Inthis paper we focus on noise reduction.This work presents the novel noise-reduction algorithmDual Energy Shifted Wavelet Denoising (DESWD), whichrenders high-quality Dual-Energy CBCT projections out ofnoisy ones. To do so, we first apply a Generalized AnscombeTransform, enabling us to use denoising methods proposedfor Gaussian noise statistics. Second, we use a 3D trans-formation to denoise all the projections at once. Finallywe exploit the inter-channel redundancy of the projectionsto create sparsity in the signal for better denoising with achannel-decorrelation step. Our simulation experiments showthat DESWD performs better than a state-of-the-art denois-ing method (BM4D) in limited photon-count imaging, whileBM4D achieves excellent results for less noisy conditions