Improved Multiscale Non-local Mean Image Denoising Algorithm

Digital image processing is a new field derived from mathematical technology andcomputer science. Image denoising is one of the widely discussed topics in this field.During the forming,transferring and memorizing of the digital images,the images areoften corrupted by different kinds of noises because of the defection of the imagingsystem,transferring medium and memorizing equipment. In the fields of patternrecognition, image analyzing and video coding, the early image denoising is veryimportant. It belongs to the image preprocessing stage from the view of imageanalysis’s flow, and belongs to the image restoration from the view of the digital imageprocessing technology. So its existence has very important status. The result of it willdirectly affect the quality and outcoming of the later processing. During the past tenyears of development, variety of classical denoising models were proposed, such as:mean filtering, median filtering,Gaussian filtering,adaptive Wiener filtering, Wavelettransform filtering, anisotropic diffusion model, total variation model, bilateral filteringand non-local means filtering algorithm etc.In this paper, we first introduce the background of noise, the basic wavelet theoryand the wavelet threshold denoising algorithm; Then we discuss two kinds of classicaldenoising model,bilateral filtering and non-local mean filtering, and analysize theiradvantages and disadvantages. Then we make further research against original NLM’sdrawbacks which contain large amount of computation and remaining noise trace afterdenoising. So we introduce fast non-local mean and an improved NLM, which greatlyreduce the computation time and increase the denoising effect. After that we introduce anew denoising method, which is guided image filtering algorithm. Its output image isderived by referring to a guide image, its denoised effect is comparable with NLM, butthis method use much less time than NLM, so this reflect very well the superiority ofthe algorithm. At last we present the image denoising method based on the waveletdecomposion, introduce the wavelet decomposition denoising method which based onthe two-scale and multi-scale decomposition respectively, and make the horizontalcomparison for the different denoising algorithm, then we make the longitudinalcomparison for the denoising effect of the two kinds of decomposing method whichbased on two-scale and multi-scale, and make detailed analysis combined with theexperimental data.