摘要
波动方程递推求解后得到的去运动模糊图像,复原结果比较自然,没有振铃效应和畸变,但是会出现明显的累积噪声。为了解决这一问题,本文应用Curvelet变换方法进行消噪,该变换具有高度的各向异性,而且本质上具有投影迭加而使噪声均值为0的性质。相比传统的像素级滤波和小波阈值消噪,有更理想的去噪效果。我们的实验数据和性能分析也证明了这一结论。
An image restored by a recursive solution of traveling wave equation is natural to visual interpretation, without ringing effect and distortion emerging, but accumulated noise appears on it obviously. Curvelet transform has variable anisotropy, its projections on different directions assure the noise averaging characteristics. Experimental results presented in this paper suggest that simple thresholding of the curvelet coefficients exhibits higher denoising quality.
出处
《信号处理》
CSCD
北大核心
2005年第5期539-543,共5页
Journal of Signal Processing
基金
国家自然科学基金(60372068)广东省自然科学基金(04020074)广东省计算机网络重点实验室CNLF-CN200402)资助项目
