A Study Of Method For Image Smoothing And De-noising Based On Statistic Filtering And PDEs

As imaging systems, transmission systems and recording systems in real worldcan not be perfect, digital image will be noised when it is being created, transmittedand recorded by various noises in different level. Accordingly pre-processing fornoised image is extremely important in pattern recognition, computer vision, imageanalysis and video coding. Quality of outcomes implemented by processes mentionedabove is affected heavily by image pre-processing.Compared to traditional linear filtering, not only noise can be removed butdetails of high frequencies can also be preserved by nonlinear filtering. In this way,nonlinear algorithms are used more and more because the output image can be moreclear and vivid. Median filtering is a typical nonlinear filtering. Impulse noise withinimage can be removed by median filtering, but important fine information would belost during the process. Based on median filtering, a new median filtering algorithm isintroduced. The new algorithm is proved effective by theoretical analysis andsimulation byMatlab software.Due to less consideration of characteristic of image, traditional smoothingalgorithms such as averaging filtering and Gaussian filtering have the equivalentresult of heat conduction equation which is also classified in the range of isotropicdiffusion. As a result edges of image are blurred, or even ruined during theimplementation of filtering. Noise removing and edge preserving can be achieved atthe same time by smoothing algorithms based on partial differential equation.Compared with heat conduction equation, anisotropic diffusion has a shape ofparabola. Those two ideal goals can be achieved by diffuse speed controlling,dominated by gradient. Typical algorithms such as Perona-Malik model have beenused widely in the fields like edge detection, image enhancement, image segmentationand object recognition.Combining an improved inverse anisotropic diffusion model with adaptivestatistical filtering algorithm, a new image smoothing algorithm is introduced in thispaper. Experiment results show that the new algorithm realizes the aims of edgeenhancement which is affected by inverse anisotropic diffusion. Impulse noise isremoved and the instability of inverse anisotropic diffusion is eliminated by the induction of adaptive statistical filtering. A conclusion can be draw that the newalgorithm achieved a very good effect of edge preservation and noise elimination.