The Research Of Multi-objective Parameters Estimation Based On Motion Blurred Image

Parameter estimation based on motion blurred image has always been one of the research highlights in the robot vision, the main theroy of which is that finding useful information needed through blurred features shown in the image and establishing relationship between blurred information and motion image to get the parameters of moved object in the image.At present in view of the qustion of the taken actual image, which is usually containing multiple targets and not able to estimate blurred parameters of single target of them. To resolve this question,this paper is based on local motion blurred image to multiple targets parameters estimation, the main contents include:First of all, the forming reasons, classification of blurred image degradation model is introduced, the motion blurred image degradation model and point spread function are analyzed emphatically, and the relationship between PSF and motion blurred parameters are researched.Then studying the spectrum characteristics of motion blurred image further through the analysis of PSF, and discussing the problems of existing algorithms on blurred parameters estimation from aspects of blurred angle and blurred measure.In the end, a method in combination with local standard deviation and existing algorithms was put forward to identify the motion blurred parameters of multi-objective at the same time. The motion blurred image is filtered by local standard deviation to extract the larger standard deviation blurred edge image feature blocks which are filtered and classified later, and then the motion blurred parameters of every feature blocks are obtained which are based on Radon transform and cepstrum algorithm. Concerning the problem of inaccuracy of algorithms in the case of containing noise, this paper introduces the pyramid weights to weight and average the blurred parameter to accurate measurement results further. And the performances of proposed algorithm are tested and analyzed through the simulation experiment, simulation results demonstrate that algorithm is improved the validity and accuracy.