The UDCT Coefficient Modeling And Application In Image Denoising

Uniform Discrete Curvelet Transform (UDCT) is a new multi-resolutionanalysis tool, how to establish an appropriate model, select the appropriateparameter estimation algorithm according to the relationship between thecoefficients of the UDCT is the key issue to solve the applications in digital imagede-noising.This paper focuses on the coefficient modeling, parameter estimation andimplication in image de-noising. Based on the Predecessors’ research on thewavelet transform, contourlet transform to select the Generalized Gaussian modeland Hidden Markov Tree (HMT) model for the UDCT coefficients, deriving thecorresponding parameter estimation algorithm then apply it in image de-noising.The main content and innovation of this paper are as follows:Firstly, Continued fractions iterative has wide convergence domain,calculation simple, fast iterative and stable; in this article, the GeneralizedGaussian distribution’s (GGD) shape parameter is derived with Continued fractionsiterative; by comparing the moment estimation, Newton-Raphson iterative, andContinued fractions iterative, experiment’s results show that: when adding differentnoise variance, based on continued fraction iterative performs better than theNewton iteration in computational complexity and accuracy, and convergence withlittle constraints from the initial value of parameters, always converge to theoptimal value.Secondly, we further optimized the algorithm for HMT by defining thevariance and state transition matrix based on the attenuation of coefficients andcontinuity between the scales, experiment’s results show that: the training time isreduced by2/5; in the use of Similarity and Peak Signal To Noise Ratio effect asthe measurement of image de-noising, under the same conditions, the algorithm weproposed has better real-time and de-noising effect than the Wavelet HMT,Contourlet HMT, UDCT HMT algorithm.