An overview of orthonormal wavelet image denoising is discussed here. Instead of postulating a stastical model for the wavelet coefficients, denoising process is directly parameterized as a sum of elementary nonlinear process with unknown weights. Them an estimate the mean square error between original image and denoising image is minimized. A statically unbiased MSE estimate stein’s unbiased risk estimates (SURE) that depends on the noisy image alone is considered. This estimate is quadratic in unknown weights and its minimization amounts to solving a linear system of equations. The existence of this a priori estimate makes it unnecessary to device a specific statistical model for the wavelet coefficient. This estimate turns out to be more accurate as more data available which is the cost of images. Here an interscale orthonormal wavelet thresholding algorithm based on the new SURE approach is implemented. Test result shown improved PSNR compared to existing state of art procedures.