Study On Adaptive Image Denoising And Image Fusion Base On Contourlet Transform

Image denoising and image fusion are advanced topics in field of digital signalprocessing, and image processing method based on Contourlet transform is theresearch focus in the field in recent years. Contourlet transform is an very importantmeans in multi-dimension analysis which can effectively realize multi-resolution ratio,local and multi-direction unfolding of image, and can make the utmost ofgeographical substantive characteristics of image, which is of wide research prospect.This paper studies fundamental and realization methods of Contourlet transformfirstly, analyzes advantages and disadvantages of the existing image denoisingmethods, expounds selection of threshold, proposes the image denoising approachwith combination of Contourlet transform with Bayes threshold, and makesexperimental simulations for denoising of wavelet threshold as well as of Bayesthreshold with Contourlet transform, with simulation results given. On this basis, animage denoising method of adaptive threshold based on nonsubsampled Contourlettransform is further proposed. The method firstly makes a nonsubsampled Contourlettransform for noise image to obtain factors of subbands in all directions to all scales,then adaptively adjust the denoising threshold according to the factors of subbands inall directions to all scales as well as energy of theses factors of subbands.Experimental results show that, besides retainment of margin details of image, thisapproach can enhance PSNR value of image and reduce phenomenon of Gibbs,compared with denoising of wavelet threshold and denoising of adaptive threshold ofContourlet.Multifocus image fusion is an important research area in the field of image fusion.In the aspect of image fusion, the paper introduces its development background andfundamental, systematically studies application of Contourlet transform in multifocusimage fusion, proposes adaptive image fusion algorithm based on the provincialcharacteristics of Contourlet transform with characteristics of multiscales andmultidirections, and with this algorithm image is decomposed with Contourlettransform, different fusion rules are used based on characteristics of each part ofimage. Aiming at low-frequency part, weighted average is used to retain details andmargin effectively, and aiming at characteristics of factors of high-frequency parts,different fusion methods are selected adaptively. Experiment testifies that the proposed algorithm is superior to the algorithm of multifocus imagefusion, withadvantage that compositive image is clear and details are plentiful.