【作者】 张旭明； 熊有伦； 尹周平；

【Author】 Xuming Zhang~1,Youlun Xiong~2,Zhouping Yin~1 (1 School of life science and technology,Huazhong University of Science and Technology,Hubei,Wuhan 430074; 2 School of Mechanical Science and Enginnering,Huazhong University of Science and Technology,Hubei,Wuhan 430074)

【机构】 华中科技大学生命与技术学院； 华中科技大学机械科学与工程学院；

【摘要】 为实现噪声污染运动模糊图像点扩散函数精确估计,提出了基于频谱特征的识别算法和改进微分自相关算法。基于频谱特征的识别算法要先对图像进行去噪处理,然后对滤波图像进行两次傅立叶变换,随后通过对频谱图进行模糊增强、轮廓跟踪和轮廓分析等处理手段,实现模糊方向的估计;改进微分自相关算法,先对去噪后的运动模糊图像在特定方向上进行微分运算,然后利用微分自相关图像建立点扩散函数鉴别曲线,通过在曲线上寻找具有极小值的负尖峰位置来实现模糊长度估计。实验结果表明:对受噪声污染的模糊图像,提出的模糊参数识别算法可实现模糊方向和模糊长度的有效估计,估计精度优于其他算法。更多还原

【Abstract】 To realize accurate estimation of point spread funcition(PSF) of motion blurred images corrupted by noise,spectrum characteristics based recognition (SCR) algorithm and improved derivative autocorrelation(IDA) algorithm are proposed.In the SCR algorithm,the blurred images are de-noised firstly and then FFT is implemented for the filtered images twice.For the resultant spectrum images, such technologies as fuzzy enhancement,contour tracking and contour analysis are adopted to estimate the blur direction.In the IDA algorithm,the derivative operation is implemented for the de-noised motion blurred images in specific directions. The derivative autocorrelation images are used to acquire point spread func- tion identification curves,on which the position of minus peak with minimum value is searched to estimate the blur length.Experimental results show that the SCR algorithm and IDA algorithm can realize the estimation of blur direction and blur length for the corrupted motion blurred images with higher accuracy than the compared algorithms.更多还原