显著边缘直接引导的动态模糊图像盲复原方法

Image blind restoration algorithm based on obvious edge feature

针对主动视觉中运动载体和云台抖动导致图像模糊的问题,提出一种显著边缘直接引导的动态模糊图像盲复原方法。为了克服经典Canny算子只能对边缘进行标注的缺陷,设计了图像边缘搜索-增强算法,并基于此提出了仅依据图像中显著边缘邻域完成点扩散函数优化估计的方法;为解决传统残差图像方法不能适应预测图像错位的问题,提出图像解模糊的递归残差修正算法来抑制复原图像波动;设计了图像复原的分层迭代处理流程,保证模糊图像复原的可靠性和精度。理想光照条件下的仿真表明,算法能准确恢复模糊图像的边缘,同时抑制复原图像中的波动。非理想光照条件下的仿真也验证了算法对环境有良好的适应性,具有较高的实际应用价值。

For the problem of the imaging blurring caused by motion of the vehicle or vibration of the active vision plat in vision measurement,a new algorithm of image blind restoration is proposed by obvious edge feature. Firstly,an algorithm for searching and enhancing image edges is designed to overcome the drawbacks of the Canny algorithm. Further,a method of optimum estimation for the PSF is proposed only based on the neighborhood area of the obvious edges. Secondly,a recurrence residual algorithm is proposed for image deblurring,which can restrain the ringing of the restored image caused by obvious edges. Lastly,a multi-layering with iteration flow path is designed for guaranteeing the accuracy and reliability of the blind restoration. Experiments demonstrate that the method can reconstruct images and suppress the ringing effectively. The experiment under un-ideal illumination condition also shows the good adaptability against environment. This image restoration new algorithm can be efficient in practical application.