一种改进的运动模糊图像参数准确估计方法

Improved approach to motion blur parameters identification

模糊角度(方向)和模糊长度是对运动模糊图像进行恢复处理的两个关键参数.针对通过运动模糊图像频谱图像的中心亮条纹获得模糊角度存在角度检测精度不高、角度检测范围受限制等缺陷,在分析Radon变换的数学原理和十字亮纹形成原因及其造成干扰的基础上,根据频谱图像的特点,提出采用模糊频谱图像的暗条纹获得运动模糊角度的方法:在获得频谱图像暗条纹的二值图像后,通过骨架化变换,缩小暗条纹宽度,再经过Radon变换获得模糊角度和模糊长度.实验结果表明,所提出的方法能避开频谱图像中十字亮线的干扰,提高模糊角度检测精度和稳定度,扩大角度检测范围,模糊角度的检测误差小于0.5?,角度检测范围达0?～180?。

The blur direction and blur length are two key parameters to recover motion-blurred images. In view of the defects that low precision and the limitation of angle range in the method of detecting center bright stripe of Fourier spectra image. Based on the analysis of the mathematical principle of Radon transformation and the causes of cross-shaped bright line and its interference, according to the characteristics of spectral images,the improved approach of using the dark line of Fourier spectra image to identify blur direction is proposed. After the binary image of Fourier spectra image is got,the skeletonisation of mathematical morphology is employed to make the width of dark line narrowed, then the Radon transform is employed to get the blur direction and blur length. The experiment results show that the proposed method can avoid the disturbance of cross-shaped bright line, and improve the precision and stability of the blur direction detection and enlarge the range of angle detection. The detection error of blur angle is less than 0.5, the range of blur angle detection is 0?~ 180?, it works better than traditional algorithms.