Wavelet Local Donoise Reconstruction Algorithm Based on Non-separable MRA

We propose a wavelet reconstruction algorithm based on non-separable MRA In the algorithm we disposed the projection data with two channel wavelet decomposement to obtain the approximate coefficient and the detail coefficient of the image, and then after denoised these coefficient with soft-thresholding, we reconstructed the image though reverse wavelets transform. We used two-dimensional quincunx sub-lattice which is better psychovisually because it means that the perceptually least valuable component of vision is quantized first. This algorithm not could reconstructed local image by local date, but could denoised.