摘要
模糊图像的盲恢复是当前图像处理研究的热点问题.本文提出一种新的基于图像块拼接的模糊图像盲恢复算法,首先提取模糊图像的直线边界,然后选择包含最多信息边界的模糊图像块进行拼接,利用模糊核函数与这些直线信息边界的对应关系,采用交替迭代算法对完成模糊核函数估计和模糊图像的恢复.实验结果证明,与基于最大似然法的Radon变换估计算法比较,本文算法在获得相同或更好的图像恢复质量条件下,运算速度是其速度的5-50倍,而且大大降低对计算机内存的要求.
Blind image deblurring is one hot problems of image processing technology research at present problem.In this paper a new method is introduced using a mosaic image patches for kernel estimation w ith minimum computational cost in motion blurred image blind restoration.Blur kernel is estimated by edges in the blurred image,effectively constructing the Radon transform of the kernel,in order to reduce computations,image patches included a lot of blurred information w ill be chosen and put together,then an alternate blind deblurred method w ill be used.Experimental results show that our method to be 5-50 times faster than M APradon method w hile achieving equal or better quality.M oreover,our method reduces the requirements of the computer memory.
作者
郭立文
廖永忠
GUO Li-wen;LIAO Yong-zhong(Department of Computer and Software,Shanxi Institute of Technology,Xi’an 710300,China;Department of Information Science and Engineering,Hunan First Normal University,Changsha 410205,China)
出处
《小型微型计算机系统》
CSCD
北大核心
2020年第10期2225-2229,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金面上项目(61175064)资助。
关键词
图像处理
模糊图像
盲恢复
RADON变换
引导滤波
image processing
blurred image
blind restoration
Radon transform
guided filtering
