Research On Image Denoising Based On Multiscale Geometric Analysis

With the development of communication technology and network technology,as well as the rising popularity of digital electronic products, the image has becomean important carrier of access to outside information. However, the image in theprocess of collection, transmission and storage are vulnerable to the interference ofnoise which results in the decrease of image quality, and the image quality has veryimportant influence for subsequent processing, therefore it is necessary for imagenoise reduction processing, in order to obtain higher quality images.Image denoising methods are divided into spatial domain denoising andtransform domain denoising.Transform domain denoising is currently very populardenoising method, wavelet transform and multiscale geometric transformation is themajor transformation methods. Relative to wavelet transform, multiscale geometrictransformation is highly directional, can achieve better denoising effect, so multiscalegeometric transformation denoising has become the research hotspot in the field ofdenoising.This paper first studied the basic theory of multiscale geometric analysis, mainlystudied the basic theory of Curvelet transform and its realization method. Thenstudied hard threshold denoising and soft threshold denoising of Curvelet transform,and experiment with global threshold and local threshold respectively, proved andanalyzed the local threshold can better adapt image energy distribution after theCurvelet transform. Relative to global threshold denoising, local threshold denoisingcan achieve better denoising effect. On the basis of the local threshold, improved interms of denoising threshold and threshold function respectively, realized Curveletadaptive denoising and improved threshold function denoising. Finally, combinedproved Curvelet denoising with total variation denoising to realize denoising, namelythe Curvelet transform and total variation combination image denoising algorithm.Compared with the previous denoising algorithm, this method obtained the bestdenoising effect, not only improved the PSNR after denoising, and improved theimage visual effect, therefore it is an effective denoising algorithm.