Low-Dose CT Reconstruction with Multiclass Orthogonal Dictionaries

We propose a high-accuracy CT image reconstruction from low-dose X-ray projection data. A state-of-the-art method exploits dictionary learning for image patches. This method generates an overcomplete dictionary from patches of standard-dose CT images and reconstructs low-dose CT images by minimizing the sum of date fidelity and reg-ularization terms based on sparse representations with the dictionary. However, this method does not take characteristics of each patch into account, such as texture and edges. In this paper, we propose to divide all patches into several classes, and use an individual dictionary with an individual regularization parameter for each class. Moreover, for fast computation, we introduce the orthogonality for each dictionary. Since clustering collects similar patches, accuracy degradation by the orthogonality hardly occurs. Simulation shows the proposed method outperforms the state-of-the-art one in terms of accuracy and speed.