Noise removal of CBCT images using an adaptive anisotropic diffusion filter

In this work we proposed an adaptive anisotropic filtering method for removing unwanted noise information that may occur in cone beam computed tomography (CBCT) images. The data used in this study consist of 1200 different image sections obtained from 30 different patients who came to Karadeniz Technical University, Faculty of Dentistry, Department of Oral Diagnosis and Radiology Clinic for routine controls. At first, to identify 2D image sections that do not contain noise information, we measured noise levels in CBCT dataset sections using a noise level estimation method. Then, we applied different levels of noise to those noise-free images. We used anisotropic diffusion filter (Perona and Malik's filter), an automatic anisotropic filter (Tsiotsios and Petrou's method), and our adaptive anisotropic filtering method to remove noise information from those images. Afterward, we obtained peak signal to noise ratio (PSNR) and mean absolute error (MAE) values derived from the results. Proposed adaptive anisotropic diffusion filter seems to be a good choice for removing noise that may occur on CBCT image sections.