基于频谱分析的运动模糊图像参数检测

Parameter detection method of motion blur images based on spectrum analysis

摘要:

为了提高运动模糊图像的运动方向和运动像素的检测精确度，本文提出了基于频谱分析的运动模糊图像参数识别的改进算法，从频谱分析的角度出发推导了暗条纹宽度、图像大小、条纹倾斜角度与运动尺度、运动像素之间的关系；采用二次傅里叶变换的方法细化亮条纹，分析并简化了条纹倾斜角与运动方向角之间的关系；分析了频谱中十字亮线产生的原因及位置，提出了去除十字亮线的新方法；然后，对处理后的频谱进行Radon变换，根据变换结果得出模糊图像的运动方向和运动像素。实验证明，该算法能够检测出不同大小模糊图像的运动方向和运动像素，检测误差控制在0.5°、2个像素的范围之内。

Abstract:

A modified algorithm for parameter identification of moving blur images is proposed based on the spectrum analysis in order to improve detecting quality of the direction and pixel for moving objects in blurred image. The relationships among dark fringe width, image size, inclination angle and kinematic scale, kinematic pixel are derived. The twice Fourier transform is applied to meet refinement of bright fringe. And then the relationship between inclination angle of fringe and moving angle is analyzed. A new algorithm for eliminating the cross bright line is proposed by analyzing its generation reasons and position. Radon transform is applied to the processed image, and the moving direction and the moving pixel are obtained according to the transform results. Simulation results show that the proposed algorithm can detect the moving direction and pixel of different sizes of blur images with a detection error under 0.5° and two pixels.