Filtragem de sinograma tomográfico afetado por ruído Poisson utilizando wavelets anisotrópicas

Abstract

The diagnostic imaging by computed tomography has become popular in recent decades. As the image acquisition method is effected by exposing the human body to X-ray dose (ionizing radiation), the frequent realization of diagnoses results in the accumulation of radiation, and may be a significant factor in the emergence of cancer diseases. This problem has motivated researchers to develop methods to reduce the dose of X-rays so that the image quality can be maintained. A common methodology involves applying algorithms for filtering noisy tomographic images that can be acquired under low dose of X-rays. This process may occur in the reconstructed image domain or tomographic projections domain. This work proposes a methodology for applying filtering Poisson noise present in tomographic sinogram, which uses the main anisotropic wavelets studied by the scientific community, which are: Curvelet, Contourlet, Shearlet. These wavelets apply the concept of multiscale and multidirectional analysis in a multidimensional signal, and may be advantageous in two-dimensional analysis of sinogram, preserving better image details in a noise reduction process compared to orthogonal wavelets.