A Fast and Accurate Dental Micro-Ct Image Denoising Based on Total Variation Modeling

Abstract

Quantitative evaluation of mineral density of carious dental lesion is one of the major aims in cariology investigations particularly in the study of caries remineralization. Nowadays X-ray micro computed tomography (Micro-CT) is used as a well-known modality for this purpose. However, the produced Micro-CT images are affected by substantial noise. To address this issue, we propose a new approach for de-noising dental Micro-CT images based on total variation (TV) modeling. The idea of applying this method traces back to the structural features of a tooth, and almost nontextural nature of noise-free images. So, using TV we intend to separate texture from cartoon which results in major reduction of the noise in Micro-CT dental images. Our simulation results on a dataset of 51 teeth of size 1000x1000 showed that our method outperforms BM3D method, currently one of the state-of-the-art de-noising methods, in terms of Contrast-to-Noise Ratio (123.02 +/- 11.29 vs. 96.79 +/- 6.87) while Edge Preservation Indexes are the same.