Some Kinds Of Digital Image Reconstruction Algorithms

Images are most important information source for human beings,but they are often interfered by noises in the process of acquisition and propagation.So denoising is an important technique of image processing.Usually,a filter or algorithm may damage some local feature of an image,since it is based on calculations of the gray values of its pixels.Therefore a good algorithm should remove image noises,and keep the image features as well.In this article we recall some classical denoising algorithm and make some improvements.We recall two algorithms based on discrete Fourier transformation and discrete wavelet transformation respectively,and then reconstruct some noised images by discrete Fourier transformation and discrete wavelet transformation with three different wavelet functions.Moreover,we introduce a denoising algorithm based on total variation model,and give three numerical iterative methods.Finally,a spatial domain filter based on the non-local mean is proposed,and a rotation-invariant similarity measure is given.We compare their performances by experiments,and conclude that the non-local means algorithm with rotation-invariant similarity measure is best.