Débruitage d’images médicales par modélisation stochastique spatiale.

Abstract

Ultrasound medical images de-noising is an important field that is used infinitely in image processing, where images are corrupted by multiplicative noise called speckle. Different methods and techniques should be used to remove these noises. This thesis presents a novel approach for ultrasound (US)images denoising. This is a class of Generalized Moment Method (GMM) estimators with interesting asymptotic properties for 2D GARCH modeling of wavelet coefficients. Indeed, these estimators are used to suppress noise in US images. An MMSE (Minimum Mean Square Error) method is applied to estimate the wavelet coefficients of the clear image. To judge the quality of the noise reduction procedure, a link between the noise reduction efficiency procedure and a proposed asymmetry measurement is established. Several tests were conducted to prove the performance of the proposed approach. The results obtained are compared to those of well-established image de-noising methods using the usual image quality assessment metrics and two proposed no-reference quality metrics.