摘要
退化模型点扩散函数（PSF）的准确估计是图像复原的关键。针对匀速直线运动模糊PSF参数未知的情况，提出了一种基于参考图频谱相关系数分析的模糊参数估计方法。利用模糊图像的二次傅里叶频谱图所呈现的运动模糊方向特征，将一幅清晰的参考图像以固定模糊尺度、不同模糊方向模糊处理，求取与待测模糊图像二次频谱的最大相关值预估模糊方向，方向误差可以通过定义的调节矩阵限制在小范围区间内，进而在一次频谱中准确计算模糊方向和模糊尺度。实验结果表明，该方法对任意方向、模糊尺度在10〈L≤80范网的无噪模糊图像的角度和尺度估计均无误差；对添加方差在0．001—0．01的高斯白噪声模糊图像，角度估计的最大绝对误差为3°，尺度估计的最大绝对误差为2个像素。
The accurate estimation of the point spread function （PSF） of degradation model is the key point in image restoration. Aiming at the fact that the blur PSF parameters of uniform linear motion are unknown, a blur parameter estimation method is proposed based on frequency spectrum correlation coefficient analysis with reference image. Spe- cifically, using the motion blur direction property of the blurred image twice Fourier transform spectrum, a clear refer- ence image is blurred with a fix blur length and various blur directions, and then the blur direction estimation is achieved through analyzing the spectrum correlation maximum between the blurred reference image and the image to be restored. The direction error can be limited in a small range by using a defined adjustment matrix. Furthermore, the motion blur direction and blur length can be calculated accurately by analyzing the spectrum correlation in Fourier frequency domain. The experiment results show that for noise free blurred image, there are no direction and length es- timation errors in any direction and within the blur length blurred image with variance between 0. 001 and 0.01, the maximum length absolute estimation error is 2 pixels. range of 10 〈 L ≤ 80 pixels;for the Gaussian white noise maximum direction absolute estimation error is 3° and the
出处
《仪器仪表学报》
EI
CAS
CSCD
北大核心
2014年第1期88-96,共9页
Chinese Journal of Scientific Instrument
基金
国际科技合作项目(2011DFR10480)
山西省自然科学基金(2013011016-4)资助项目
关键词
运动模糊
参数估计
频谱相关性
抗噪性
motion blur
parameter estimation
spectrum correlation
anti-noise
