空间目标图像的非凸稀疏正则化波后复原

发布时间：2018-04-29 16:00

本文选题：空间目标图像 + 波后复原　；参考：《光学精密工程》2016年04期

【摘要】：现有的空间目标图像波后处理方法多直接套用自然光学图像的复原技术,效果并不理想。本文通过分析空间目标图像的近似稀疏性和灰度值服从超拉普拉斯分布的独有特点,提出了一个采用正则化方法的非凸稀疏正则化空间目标图像复原模型。在数值计算过程中,根据交替方向乘数法将复原模型分解为两个子问题,对凸优化子问题采用快速傅里叶变换求解,对非凸优化子问题采用固定点迭代方法求解。文中设计了非凸稀疏正则化空间目标图像波后复原的完整算法流程,并针对模拟图像和真实空间目标图像进行了对比验证。结果显示:相对于最近的流行算法,提出方法的最大峰值信噪比提高了2dB,最大平均结构相似度提高了0.17,最大信息熵提高了3.85,图像清晰度提高了2.65。
[Abstract]:The image wave post-processing method of the existing space target is mostly used to directly apply the restoration technology of natural optical image, and the effect is not ideal. By analyzing the unique features of the approximate sparsity and gray value of the space target image, this paper proposes a non convex sparse regularization space target map using the regularization method. In the process of numerical calculation, the restoration model is decomposed into two sub problems according to the alternating direction multiplier method. The convex optimization subproblem is solved by fast Fourier transform, and the fixed point iteration method is used to solve the non convex optimization subproblem. The complete calculation of the wave recovery of the non convex sparse regularization space target image is designed in this paper. The results show that the maximum peak signal to noise ratio of the proposed method is improved by 2dB, the maximum average structure similarity is increased by 0.17, the maximum information entropy is increased by 3.85, and the image sharpness is increased by 2.65..

【作者单位】：信息工程大学理学院;信息工程大学地理空间信息学院;郑州升达经贸管理学院;
【基金】：国家863高技术研究发展计划资助项目(No.2012AA7032031D) 国家自然科学基金资助项目(No.11373043)
【分类号】：TP391.41

【相似文献】