静态背景中目标运动去模糊

Object Motion Deblurring in Static Background

为解决静态背景中目标运动去模糊的难题,提出了一种基于透明性的目标运动去模糊方法.首先对目标的1-D运动模糊的透明性进行研究和分析,在传统卷积模型的基础上得出目标透明性与模糊滤波之间的关系式,并利用该关系式计算出1-D模糊滤波;然后对目标的2-D运动进行分析,在2-D空间上利用透明性估计出模糊核尺寸的上边界;最后利用共轭梯度优化和置信度传播估计模糊滤波和未模糊的透明性映射图,并采用贝叶斯准则的最大后验分布的方法完成目标运动去模糊的任务.基于人工合成的图像和实际拍摄的图像的实验结果表明,该方法能够很好完成静态背景下目标运动去模糊的难题,性能优于当前技术条件下的运动去模糊方法.

To solve the object motion deblurring problem in static background,an object-motion deblurring approach based on transparency is proposed.Firstly,the transparency of object 1-D motion is studied and analyzed,the relationship equation between object transparency and blurred filtering is gained on the basis of conventional convolution model,and the 1-D blur filter is accurately computed with that equation.And then,the object 2-D motion is analyzed,and the upper bound for the size of the 2-D motion blur filter is estimated with transparency in the 2-D space.Finally,the unblurred transparency map and blurred filtering are estimated using conjugate gradient optimization and belief propagation,and the task of object motion deblurring is accomplished with a maximum a posterior approach of Bayesian rule.The experimental results based on both synthesized images and real images show that the proposed approach can achieve the object motion deblurring problem very well in the static background,and the performance is better than current state of the art approaches for motion deblurring.