Digital Image Denoising Algorithm Based On Multiscale Transform

With the rapid development of technology,digital images have become a new carrier of human information exchange.During the process of generating and transmitting noise is inevitable.These noise not only affect the visual effects of the image,and even change the image of the content and quality,which causes great disruption to the subsequent digital image processing operations.Image denoising is the process of reconstructing the original image by removing unwanted noise from a corrupted image.It is designed to suppress the noise,while preserving as many image structures and details as possible.A large number of simulation experiments show that the larger the number of modulus models in transform domain,the more image information is included.Therefore,based on this theory,the following two image denoising algorithms based on transform domain are proposed:1.Image Denoising Algorithm using non-subsample Contourlet domain Based on Polar Harmonic Transforms.Firstly,the image was decomposed in high-frequency and low-frequency subbands of frequency and orientation responses using the nonsubsampled contourlet transform.Then those subbands are divided into blocks by the sub-band fixed-size window and obtained by calculating the similarity of each pixel block of PHT.Secondly,the similarity calculated as the weight value of the pixel to modify the pixel gray value.Finally,the modified nonsubsampled contourlet coefficients were transformed back into original domain to get the denoised image.This method not only preserves the advantages of traditional nonlocal mean denoising methods,but also incorporates non-subsample Contourlet conversion with good frequency selectivity and regularity.It transforms the image denoising process from the airspace into the transform domain and calculates the similarity with PHT.The experimental results also show that the algorithm has good denoising performance and strong robustness.2.Image denoising method using PDTPFB based on Proximal Classifier with Consistency.Firstly,the denoising method based on PDTDFB transforms the image contents into low and high subbands.Then it use the appropriate threshold to determine the binary table for each subband and constructs the eigenvector with the space rule.Secondly,the denoising method use PCC for training thresholds.Through the training,the hresholds are divided into different orientations and scales adaptively.The third step is to remove noise from every subband.Finally,the image is reconstructed by inverse PDTDFB transforming.This method introduces PCC,which improves the problem of high time complexity and poor classification of support vector machines.And through PDTDFB transform,it can achieve a higher degree of redundancy in the higher resolution and multi-resolution resolution.The experiments show that the algorithm improves the edge protection ability and is good at denoising.3.The algorithm is applied to the denoising system based on Matlab platform.And this system,which is simple operation,full-featured advantages,can make the proposed algorithm be more easily applied to the actual de-noising work.