Analysis of Noise Properties in Dental Images

Images used in medicine are often noisy. Noise might originate from different factors and it usually degrades image quality leading to less reliable diagnostics. There are stages and the corresponding methods of image processing for which it is extremely desired to know image characteristics to take them into account. In particular, this relates to dental images for which noise is clearly seen and its properties can differ from traditional assumptions if nonlinear operations are carried out to improve visual appearance of acquired images. In this paper, we apply several known (earlier designed) approaches to automatic (blind) estimation of noise statistical and spectral characteristics. It is shown that noise in dental images is spatially correlated and signal dependent with specific dependence. The obtained estimates of noise parameters demonstrate that signal-dependent component is prevailing and the term proportional to squared intensity is present. This can be a serious problem for many image processing techniques.