Application Research On Contourlet Transform In Image Denoising

Image denoising is a hot topic in image processing.The idea of image denoising algorithms is to eliminate noise and remain the original image information as much as possible.Contourlet transform is a very effective method of multiscale geometric analysis (MGA), which can overcome the disadvantages of wavelet transform. It has multi-resolution, localized and multi-direction and other excellent features. The transform also can more effectively capture geometric structure of images. Therefore, Contourlet transform has broad application prospect in image denosing.Firstly, the theory and the realization algorithm of Contourlet transform was studied. Then application of Contourlet transform in image denoising was further discussed. The contents were as follows:1. An algorithm based on Contourlet transform which improve threshold function in image denoising was proposed. Compared uniform threshold function with traditional threshold function, a new threshold function was constructed. The new threshold function partly improved over-strangling of uniform threshold and can resolve the pseudo Gibbs phenomenon of hard threshold.2. An image denoising algorithm was Proposed that based on the combination of Contourlet transform adaptive threshold and Cycle Spinning. Based on analysis of Contourlet coefficient, the adaptive threshold on energy difference of different scales and orientations was obtained. Cycle spinning was used to denoise. The simulation showed that better denoising effect was got and PSNR value was improved after denoising. At the same time, the details of the image was reserved very well.3. Under mixture noise background, an denosing approach based on combination of median filtering and Contourlet transform was put forward. Firstly, the noise point of impulse noise was detected by median filtering and disposed. Then Gaussian noise was handled using Contourlet transform. The experiment results showed that the approach can effectively filter impulse noise and Gaussian noise, also can improve PSNR value. The most important is that the details of images were kept and improve visual effects.