【摘要】 针对目前运动模糊图像盲复原算法对图像边缘中拐角结构复原不佳这一问题,提出了一种以各向异性总变分为图像和模糊核正则项的遥感图像盲复原方法。为便于进行数值计算,采用交叉算法和分裂布雷格曼迭代导出了本文提出的盲复原方法的迭代公式。实验结果表明,与基于同向异性总变分的盲复原方法以及基于小波框架的盲复原方法相比,本文方法不仅能估算出较精确的点扩散函数(PSF,point spread function),有效地去除图像的模糊效应,而且对图像边缘结构特别是拐角边缘结构的增强有着独特的优势。更多还原

【Abstract】 The edge is an important feature of images,which is well used in many practical engineering fields,such as edge detection,object recognition and object tracking.However,the images often suffer from the blur effect due to motion between the object and the camera during the image acquisition process.Numerous methods have been proposed for this problem,and the total variation based methods are the most popular and efficient ones.But,images deblurred by using total variation based methods often suffer from serious stair-casing effects,which would damage the quality of the corners on the edge of images.In this paper,to protect the corners of the objects in images,an anisotropic total variation based regularization method is proposed for blind image deblurring,which regularizes both image and blur kernel by using anisotropic total variation instead of total variation.Besides,efficient iterative formula based on alternative method and split Bregman iteration is derived to apply the improved method in numerical experiments.Moreover,the results reveal that compared with the total variation based methods and framelet methods,the presented method can estimate the precise point spread function(PSF)and remove the blurring as well as protecting the details of corners of objects.更多还原