Research And Application Of Image Denoising Method Based On Curvelet Transform

The images usually bring different kinds of noises in process of receiving, coding and transmission. It is necessary to eliminate noises for subsequent higher disposal. However, traditional methods damnify edge characteristics. Wavelet has the good capability of expressing the unidimensional smooth partition signal, which is not suitable to express two-dimensional signal. Although the noise is eliminated and detailed edge information of the image is lost, which makes the image illegible. How to find a kind of method that not only is able to reduce the noise effectively but also can retain edge information of the image well is always being pursued by people.Curvelet transform is one kind of new multiscale transform after 1999 that is based on wavelet transform, whose structural elements include the parameters of dimension and location, and orientation parameter more, which let Curvelet transform has good orientation characteristic. Therefore, Curvelet transform is superior to wavelet in the expression of image edge, such as geometry characteristic of curve and beeline, which has already obtained good research results in image denoising. This paper puts forward a improved method based on Curvelet transform because certain regions of the image have "the ringing" and the radiated stripe after Curvelet transform.The experimental result indicated that, the improved Curvelet transform has a abroad future for eliminating the noise of images. It suits not only the ordinary visual image, but also remote sensing image. The remote sensing image, such as SPOT image, Landsat TM multi-spectra image and so on, can all make use of the improved method of Curvelet transform that this paper is proposed to eliminating noises.