The Study Of Image Denoising Based On PCA In Curvelet Domain

Traditional wavelet analysis can only express a point-like singularity in one or two-dimensional signals, it couldn’t show a good representation at the edge of the curve characteristics for two-dimensional digital image. The curvelet analysis is a combination of multi-resolution analysis and directional filtering methods, which has a significant advantage in expressing the directional characteristics in two-dimensional signal and has the characteristics of anisotropy, multi-directional characteristics except for the multi-resolution wavelet analysis, time-frequency localized compared to the wavelet analysis. Curvelet can more effectively capture the texture information in the natural image, due to its coefficient, in a particular direction, has better clustering, it can be more effectively carried out the two-dimensional digital signal sparse representation.This paper discussed the application of Curvelet theory in reducing the image noise and researched the traditional threshold denoise based on wavelet, ridgelet and curvelet. Finally, a new algorithm-based on the Curvelet transform the image PCA noise reduction algorithm has been proposed. In an overview of the research background and significance of image noise reduction. To reduce the noise of image by using traditional threshold noise reduction algorithm has a disadvantage of poor adaptive capacity and introduce some undesirable affection such as pseudo-Gibbs. This article combined the Cycle Spinning with image PCA noise reduction based on curvelet transformation,which improved the peak signal-to-noise ratio and inhibited the Pseudo-Gibbs phenomenon and radial stripes during reducing noise.The simulation results show that the method metioned in this paper can obviouslt improve the peak signal-to-noise ratio of the image after reducing the noise of the image and make the visual effects better.