Análise de métodos para suavização de ruídos em imagens de tomografia computadorizada multislice de baixa radiação

Abstract

The objective of this work is to identify a computational method to smooth noise in images of low radiation multislice computed tomography (MDCT). These images contains low quality, contaminated by noises, which are stochastic phenomena and impossible to predict their occurrence. In order to perform this task, some image processing techniques for noise smoothing were studied. PSNR, SNR, MSE and SSIM metrics were used to evaluate de quality of the processed images. The filters analyzed and simulated were the median, average, mode, gaussian and Wiener, from the spatial domain, and basic operators of mathematical morphology. After some simulations, it was verified that the gaussian filter technique presented superior results with an average PSNR of 25.64dB and an average SSIM of 0.76, for the best cases.