Study On Curvelet Transform And Its Application In Image Processing

Wavelet analysis has achieved great develetment in singal processing. Wavelet transform has a lot of advantages, but the wavelet applied to image processing has still limitations. Wavelet analysis has not a good expression of anisotropy of the image feature, because it uses point to describe the characteristics of image mutationrather than continuous curve. wavelet transform is not suitable for that singular line of two-dimensional image signals. Curvelet transform is a new method of multiscale transform, which was developed from wavelet and ridgelet transform. Curvelet transform is especially suitable for representing two-dimensional signals with linear or hyperlane singularities, and it is highly directional. Curvelet transform has high anisotropy,better approximation to curve and provides more directional information of image, so it has a better description to the edge,details and curve characteristics of image. Curvelet transform can more sparsely express image so that it makes signal energy more concentration, so it provides a description of a more favourable data expression to image. Therefore, in recent years the research of Curvelet transform is a hot spot in the field of image processing.On the basis of Curvelet theory, this paper studies the concrete realization Curvelet transform and the coefficient distribution of image using Curvelet transform, and studies Curvelet transformation in image processing applications. Concrete work as follows:(1)According to the defects of soft thresholding and hard thresholding image denoising methods, the method between soft and hard thresholding image denoising and a structural threhold function denoising approach in curvelet domain are proposed. The curvelet transform coefficients in different subbands are filtered with adaptive thresholds. Experiment results show that the methods yield denoised images with higher quality recovery of edges,linear and curvilinear features. They are capable of achieving the higher peak signal-to-noise ratio (PSNR) and giving better visual quality.(2)A algorithm for noisy image adaptive enhancementbased on the Curvelet transform is proposed, which aims at minimizing image noise and highlighting image details. The proposed algorithm can adaptively adjust the ranges of the restrained noise and the intensity of the signal to different Curvelet scales. The results of experiment show that the algorithm performs better in restraining noise,protecting image edges and highlighting image properties. (3)Two new image fusion algorithms are proposed based on the Curvelet transform. Firstly, the source images are decomposed using Curvelet transform, then the coarse scale coefficients and fine scales coefficients are fused with different fusion regulars, and finally the fused coefficients are reconstructed to obtain fusion results. The results of experiment indicate that these methods yields fusion image with better visual quality and better objective index in entropy, mean square error, difference coefficient, correlation coefficient and so on.