Image Denoising And Enhancement Based On Curvelet Transform

Wavelet analysis has achieved great development in signal processing, which profits from its time-frequency analysis. But wavelet analysis can only deal with the situation where points are singularity, it can't express features of an image along the edge effectively. Curvelet transform provides a new multi-resolution analysis method with various features, which represents edges better than wavelets and is therefore well suited for Multi-scale edge. Curvelet transform has high anisotropy, better description of edges and detail information of images. It has great potential for image processing.In this paper, firstly the background and significance in image denoising and enhanceme-nt are addressed. Then the classical methods and achievements are overviewed and summariz-ed. The basic theory of Curvelet transform, which belongs to the Multi-scale Gemetrical Anal-ysis (MGA), is discussed in depth. Two classical fast discrete Curvelet transform algorithms are analyzed. Based on these previous researches, the new algorithm of image denoising and enhancement are developed.In image denoising, a variety of image denoising algorithms are studied. According to the drawback of soft thresholding and hard thresholding denoising algorithms, a new method that compromise soft and hard thresholding approach is proposed. The Curvelets coefficients in different subbands are filtered with adaptive thresholds. We used the Peak Signal to Noise Ratio(PSNR) and subjective visual quality to estimate the image denoising quality. Compared with hard threshold denoising and soft threshold denoising based on wavelets domain, and other threshold regulation algorithms based on Curvelets transform domain, the proposed method can preserve more edges and texture of the reconstruction images. Experimental results show that our algorithm improves PSNR for 1.5dB than traditional method.In fog-degraded color images enhancement, firstly two classical enhancement algorithms—histogram equalization and multi-scale Retinex are reviewed. Then a new enhancement method based on vanishing point detection and second generation Curvelet transform is proposed.The experimental results show that the overall image enhancement is achieved and the image contrast is improved, which is better than the prior two methods.