摘要
针对离焦模糊图像高斯核难以快速精确估计的问题,提出了局部熵和参数校正相结合的辨识算法.该算法首先对模糊图像进行局部熵滤波提取图像阶跃边缘模糊区域,然后对此区域进行两次人为离焦模糊,根据梯度比估算高斯模糊核,在计算中提出参数校正的方法减少了量化误差,实验结果表明该算法在模糊核较小时能够精确快速地定位阶跃边缘,并提高了高斯模糊核的识别精度和识别效率,为模糊图像复原提供了较为精确的退化函数.
The algorithm combining local entropy and parameter correction is produced aiming at the question that it is difficult to estimate the Gaussian Kernel quickly and correctly in blind restoration of defocused image.Firstly,the gray level information of blur image is extracted by local entropy filter and twice artificial defocus blur steps are carried out on this blurred domain.After this step,the Gradient radio is compused to estimate Gaussian Kernel.So aiming at reducing quantization error,parameter correction method is put forward.The experiment results show that the step edges can be located quickly and accurately when the Kernel is smaller and recognition accuracy and recognition efficiency of Kernel are increased which can provide more accurate degenerate function for blurred image restortion
出处
《河南师范大学学报（自然科学版）》
CAS
北大核心
2013年第4期149-152,共4页
Journal of Henan Normal University(Natural Science Edition)
基金
国家自然科学基金(U1204609)
河南省教育厅基础与前沿项目(2010A520027)
河南省教育厅基础与前沿项目(2011A520026)
关键词
离焦模糊
高斯核
局部熵
量化误差
阶跃边缘
defocuse
Gaussian Kernel
local entropy
quantizstion error
step edge
