利用频谱特性鉴别运动模糊方向  

Identification of the motion blur direction using spectral property

作　　者：梁宛玉[1] 孙权森[1] 夏德深[1] 

机构地区：[1]南京理工大学计算机科学与技术学院,南京210094

出　　处：《中国图象图形学报》2011年第7期1164-1169,共6页Journal of Image and Graphics

基　　金：国家自然科学基金项目(60773172);江苏省自然科学基金项目(BK2008411);教育部博士学科点基金项目(200802880017)

摘　　要：点扩展函数(PSF)的精确估计是运动模糊图像恢复的关键。匀速直线运动模糊的PSF参数主要由模糊角度(方向)和模糊尺度两个参数组成,然而模糊角度的估计又是重中之重。针对R Lokhande等人提出的霍夫变换的运动模糊方向估计法进行改进,通过增加边缘检测等预处理步骤,利用霍夫变换检测直线,并利用改进的霍夫变换峰值提取方法来估计角度值。实验表明,该方法能够得到比原方法更精确的角度值,并且具有抗噪能力强、鉴别精度高的优点。The accurate estimation of the point spread function （PSF） is very important in restoration of motion-blurred image. The uniform linear motion blur PSF is composed of two parameters, namely blur angle（direction） and blur length. However, estimation of the motion-blurred angle is the most important. The motion-blurred direction based on Hough transform, which is proposed by R. Lokhande et al, is improved. The improvement adopts more pretreatments such as edge detection, detect the straight line by Hough transform, and takes advantage of the improved Hough transform peak extraction method to estimate the value of the angle. Experiments show that the method can be more accurate than the original method of angle values, and has anti-noise ability, identification of high precision.

关 键 词：模糊图像 点扩展函数(PSF) 边缘检测 霍夫变换 峰值提取 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]