基于双树复数小波变换的多帧迭代盲解卷积算法

摘要

This paper presents a multiframe IBD algorithm based on the Dual-Tree Complex Wavelet Transform (DT CWT). The convergence and uniqueness of the traditional IBD algorithms are both uncertain,and the reconstructed result is ver y sensitive to the initial estimation. Zhulina proposed a multiframe iterative blind deconvolution, which has a simple principle and can tackle with degraded images caused by different types of PSF. But it takes a long time to achieve convergence and needs many observations to get a preferable result. Furthermore,the algorithm is only suitable for images of the high Signal to Noise Ratio (SNR) because its sensitivity to the noises. Considering the multi-scale and good directional selectivity of the DT CWT,this paper presents the multi-frame IBD algorithm based on DT CWT. The algorithm works efficiently and it can clearly reconstruct badly degraded images blurred by noises. The experiment results prove the superiority of this algorithm.%提出一种基于双树复数小波变换的多帧迭代盲解卷积(IBD)算法.传统的单帧IBD算法收敛性和唯一性不确定,而且复原结果对初始估计很敏感.Zhulina提出的多帧迭代盲解卷积算法,其算法原理简单,并能处理各种不同类型PSF引起的图像降质;但是该算法收敛缓慢,并且只适合于处理高信噪比图像.本文基于双树复数小波变换的多尺度多方向特性,提出了一种基于双树复数小波变换的多帧IBD图像复原算法.本文算法运算速度快,且对噪声污染严重模糊图像恢复效果较好,观测数据实验结果证明了本文算法的优越性.