【摘要】 新型Contourlet变换取代第二代Bandelet变换中的二维小波变换,获得了一种新的图像变换方法——CBlet变换。其中新型Contourlet捕获图像边缘的几何结构,Bandelet化过程自适应地搜索到最佳的几何结构并去除它们的相关性,以获得更稀疏的图像表示。采用分层多阈值去噪方法,并且引入cycle spinning操作来减少伪Gibbs和皱波效应,提高了去噪图像质量。实验表明,无论在峰值信噪比(PSNR)还是视觉质量,基于CBlet变换的去噪方法均优于基于新型Contourlet变换的方法。更多还原

【Abstract】 A new image transform method named CBlet transform is obtained by substituting a new Contourlet for 2D wavelet transform in second-generation Bandelet transform.Thereinto,the new Contourlet transform captures the geometry of image edges,then the bandeletization procedure searches for the optimal geometry adaptively and further removes their correlation to obtain much sparser image representation.Layered multi-threshold de-noising method is adopted,and cycle spinning operation is introduced to reduce Gibbs-like and ripple phenomena and improve the quality of the de-noised image.Experiments demonstrate that the de-noising method based on CBlet transform can outperform the method based on new Contourlet transform both in terms of PSNR and visual quality.更多还原