基于信息论匀质性测度的各向异性扩散相干斑抑制算法

 针对传统各向异性扩散抑斑算法存在的均匀区域噪声平滑不充分、边缘随迭代弱化及迭代次数的确定缺乏理论指导等问题,提出了一种新的各向异性扩散抑斑算法,该算法采用信息论匀质性测度作为图像中匀质区域与边缘的鉴别因子,使扩散系数能够更准确地控制扩散强度与扩散速率,从而达到充分平滑均匀区域噪声及保护边缘的目的。基于各向异性扩散方程求解与鲁棒误差范数最小化的等效性,提出了一种各向异性扩散方程的迭代停止准则。利用合成孔径雷达图像对本文算法的抑斑和边缘保持性能进行了仿真实验验证。结果表明,本文算法在均匀区域相干斑噪声抑制、边缘保持等方面均取得了优于传统算法的效果。

Speckle Suppression Using Anisotropic Diffusion Based on Information-theoretic Heterogeneity Measurement

Current speckle reduction anisotropic diffusion algorithms cannot adequately smooth the speckle noise in homogeneous areas and make the edges blurred as the number of iteration increases. A novel speckle reduction anisotropic diffusion algorithm is proposed in this paper based on information-theoretic heterogeneity measurement which can effectively discriminate the homogeneous area and the edge area. With this useful property of the information-theoretic heterogeneity measurement, the diffusion coefficient can better control the diffusion strength and the speed to fully suppress the speckle noise and meanwhile preserve the edges. Based on the equivalence of solving the anisotropic diffusion equation and minimizing the robust error norm, a stopping iteration criterion is designed for the anisotropic diffusion equation. Synthetically speckled and real synthetic aperture radar images are utilized to test the smoothing performance and the ability to preserve edges of the proposed algorithm, and the experimental results show its efficiency and superiority over the traditional algorithms.