A comparison of projection domain noise reduction methods in low-dose dental CBCT

Dose reduction for CBCT can be achieved via using a low-mAs protocol. However, the quality of the projection image will be degraded by excessive noise. In this paper, we put forward four of the popular and state-of-the-art denoising methods, including TV minimization method, BLS-GSM, nonlocal means and BM3D into low-dose CBCT projection denoising. A projection domain processing algorithm based on Anscombe transform is proposed in this paper. We compare the performances of the four methods by different phantom and clinical datasets using a novel dental CBCT system. It is quite difficult to compare different methods quantitatively as there are a lot of parameters in each method. Here we optimized the methods and present their denoising effects. Non-local means and BM3D methods perform better in denoising and detail information preservation.