Algorithm Research About Image De-noising Based On Curvelet Transform And Partial Differential Equations

Image is the main media that people use to convey and acquire information. Image will be interfered by various noise at acquire and transmission process. Image de-noising is necessary preparations for subsequent image processing. Thus, it has the vital significance in the whole image processing. To seek a method which can reduce noise and can keep image edge, texture greatly becomes the research focus in the field of image processing.Curve-let analysis, a new direction and multi-scale analysis developed over wavelet and ridge let analysis can focus image subtle changes. As a new kind of multi-scale analysis method, it is strongly directional, namely the anisotropic properties by introducing a direction parameter, which is beneficial to efficiently say image edge. SO it is more useful in image de-noising area; Partial differential equations method of image de-noising is proceed on whole image. It can de-noise according to the characteristics and the direction of image. It won't cause fuzzy of image edge information during image de-noising. Partial differential equations de-noising models (such as P - M model and TV de-noising models) have good ability on edge maintenance, but also lost texture and cause staircase effect.Firstly, a new threshold function used in Wavelet threshold de-noising and Curve-let threshold de-noising was put forward on the basis of the traditional the soft and hard threshold function by earnest study. Experimental results show that the new function overcomes faults of the soft and hard threshold functions, the new algorithm has achieved better visual effect and higher peak value signal-to-noise ratio.Secondly, a new algorithm of image de-noising based on curve let transform and partial differential equations was put forward on the base of detail study on the curve let threshold de-noising models and P-M diffusion model and TV de-noising models. The new algorithm organic combine advantages of curve let transform and partial differential equations de-noising method organically, overcoming disadvantages of the curve let threshold de-noising, P-M diffusion model and TV de-noising models .It can keep the image edge, texture, and other important information preferably. Experimental results show that the new method has higher PSNR and better visual effect compared with curve let threshold de-noising, P-M diffusion model and TV de-noising models.