This paper proposes a spatially adaptive wavelet thresholding method for image denoising based on context modeling. The method models each wavelet coefficient as a random variable from a generalized Gaussian distribution with unknown parameters. Context modeling is used to estimate these parameters for each coefficient, enabling the thresholding strategy to adapt to local image characteristics. This approach improves wavelet thresholding performance by incorporating local information about smooth or edge regions. The method is extended to overcomplete wavelet expansions, which yield better results than orthogonal transforms. Experimental results show that the spatially adaptive wavelet thresholding method achieves significantly superior image quality and lower mean squared error (MSE) compared to uniform thresholding. The method uses context modeling to estimate the standard deviation of the signal, which is then used to adaptively determine the threshold. The algorithm is applied to both orthogonal and overcomplete wavelet expansions, with the overcomplete expansion providing better results. The method is effective at removing noise in smooth regions while preserving edge and texture details. The paper also discusses alternative approaches, including different variance estimation methods and heteroscedasticity modeling, but the proposed method is shown to be effective and superior in denoising performance. The results demonstrate that spatially adaptive thresholding in an overcomplete expansion significantly improves denoising quality compared to uniform thresholding.This paper proposes a spatially adaptive wavelet thresholding method for image denoising based on context modeling. The method models each wavelet coefficient as a random variable from a generalized Gaussian distribution with unknown parameters. Context modeling is used to estimate these parameters for each coefficient, enabling the thresholding strategy to adapt to local image characteristics. This approach improves wavelet thresholding performance by incorporating local information about smooth or edge regions. The method is extended to overcomplete wavelet expansions, which yield better results than orthogonal transforms. Experimental results show that the spatially adaptive wavelet thresholding method achieves significantly superior image quality and lower mean squared error (MSE) compared to uniform thresholding. The method uses context modeling to estimate the standard deviation of the signal, which is then used to adaptively determine the threshold. The algorithm is applied to both orthogonal and overcomplete wavelet expansions, with the overcomplete expansion providing better results. The method is effective at removing noise in smooth regions while preserving edge and texture details. The paper also discusses alternative approaches, including different variance estimation methods and heteroscedasticity modeling, but the proposed method is shown to be effective and superior in denoising performance. The results demonstrate that spatially adaptive thresholding in an overcomplete expansion significantly improves denoising quality compared to uniform thresholding.