The Application Of Contourlet Transform In Image Denoise And Edge Detection

The noise is the key factor to affect the quality of the image and disturbs people to obtain information, so it is necessary to eliminate noise before analyzing and applying images. And the information of contours and edges is the important research contents in digital image processing and machine visualization.Combined with contourlet and traditional image denoise on wavelet, this thesis focuses on the new threshold and new method of correlation. And according to image edge detection, on the basis of anisotropic high pass filter, it proposes a new method to apply contourlet to image edge detection. The main research of this thesis is as follows:Since the coefficients of signal are larger than the coefficients of noise, we propose a new adaptive Bayesian threshold which decides by the value of the coefficients of wavelet. This new threshold can eliminate the noise more effectively than the effect of traditional Baysian threshold and it can realize easily and keep the edges and detail well. Since the contourlet transform has the same character as the wavelet transform, and it can represent the contour and edge of image with fewer coefficients, so it is used in the process of image denoise. In the article, the contourlet and wavelet transform are both used to decompose the image and then together with the adaptive threshold to denoise the noisy image. The results show that the contourlet can keep the contour and edge better than wavelet. At the same time the adaptive threshold can adjust with the value of the coefficients, so this algorithm can distinguish image from noise better.In traditional image denoise on spatial correlation at several adjacent scales, the supposed image coefficients are picked out to reconstruct the denoised image. But the algorithm can not pick up the smaller coefficients of image, so the reconstructed image is blurred. It proposes a new algorithm to ensure smaller image coefficients can be picked up so that the reconstructed image can keep detail and edge well while suppressing noise. It also provides a new method on image edge detection which uses the DFB to detect the edge of image. In DFB decompostion, the multi-direction sub images are seen as the anisotropic filter's long axes rotating to different angles. When the anisotropic filter's long axes is in accord with the edge, the filtering performance becomes best,and the gradient amplitude is the greatest, while the angle of edge's orientation and the filter's long axes becomes larger, the filtering performance becomes worse,and the gradient amplitude becomes smaller. So through using suitable algorithms, it can obtain the edge by synthesizing the multi-direction sub images. The results show that this method is fast and simple and can also detect the edge effectively.Since applying DFB in image edge detection also has the defects such as there is only on level and there exists low frequency image. The contourlet has the benefits of DFB and also can separated the images into multi-level, it provides a new method to use contourlet transform in image edge detection which can dectect the edge in different levels and has the character of multi-direction. It shows the contourlet is a better tool to be applied in image edge detection.