基于稀疏先验的空间目标图像盲反演方法

Sparse Prior-based Space Objects Image Blind Inversion Algorithm

针对图像盲反演算法未考虑空间目标图像自身特性,致使对空间目标图像细节信息恢复不理想、重构图像中易产生边界伪像等不足之处,提出了一种基于稀疏表示的联合稀疏先验约束盲反演算法.首先,结合空间目标图像梯度的稀疏特性,采用图像梯度的L 0范数提取有利于模糊核估计的图像显著边缘信息;其次,采用L p范数和L 0范数对图像的梯度分布和空间域进行稀疏约束,以保证反演图像的像素点间具有显著的对比度,同时保证图像中包含边缘和纹理等细节信息;最后,采用拉普拉斯分布先验对模糊核进行约束,以保证模糊核的稀疏特性.采取交替迭代策略对所提出的模型进行优化求解,从而得到模糊核和空间目标图像的估计值.实验结果表明,相比于几种具有代表性的盲反演算法,提出的方法能估计出更准确的模糊核,对图像边缘和纹理等细节信息具有更好的恢复能力,在主观评价和客观评价方面均取得了较好的反演性能.

Aiming at the unsatisfactory restoration of the detail information such as boundary artifacts for the conventional blind image inversion algorithm does not consider the characteristics of the spatial target image itself,a joint sparse prior constraint blind inversion algorithm based on sparse representation is proposed.Firstly,according to the sparse feature of space object image gradient,the L 0 norm of image gradient is used to extract the salient edge information of image which is beneficial to blur kernel estimation.Secondly,the L p norm and L 0 norm are used to constrain the gradient distribution and space domain of image,so as to ensure the significant contrast between the pixels of the inverted image and the inclusion of edges and textures in the image.Finally,Laplacian distribution priori is used to constrain the blur kernels in order to ensure the sparseness of the blur kernels.An alternative iteration strategy is adopted to optimize the proposed model,and then the estimated values of the blur core and the space target image are obtained.The experimental results show that,compared with several representative blind inversion algorithms,the proposed method can estimate more accurate blur kernels,and has better restoring ability to image edge and texture details,and achieves better inversion results under both subjective and objective evaluation.