【摘要】 为了提高图像处理的实时性,本文针对作者先前提出的循环迭代图像复原算法迭代次数多、收敛速度慢等问题,提出了一种基于线性搜索的加速迭代技术.该加速技术利用泰勒级数展开式根据当前次的迭代结果对下次迭代数据进行预测,合理地跳过原有的一些迭代点,从而加快算法收敛.在达到相同复原效果条件下,减少了迭代次数,提高了算法的实时性.本文对收敛加速技术进行了理论分析和推导,在微机上进行了一系列的对比实验,实验结果表明,本文给出的加速技术效果显著,获取了加速因子的有效取值等一些有用的实验参数,算法具有实用价值.更多还原

【Abstract】 Aiming at the problems that our previous proposed the circulation iterative restoration algorithm needs many iterative times and it converges very slowly,an iterative accelerating technique based on linear search was presented here to decrease the time of image processing.By adopting the Taylor series of expansion items,the acceleration technique estimated in advance the next iterative result based on the current iterative result,and skiped some previous iterative points.In this way,it made the algorithm converge more quickly.On condition that the almost same restoration result as the previous restoration algorithm was obtained,the required iterative times were reduced and the real-time processing was heightened.Theoretical analysis and derivation of the accelerating convergence technique were presented.A series of comparison experiments had been made on PC.Our results show that the proposed acceleration technique has a significant effect.Some useful experimental parameters such as valid accelerating factor are obtained.The proposed restoration algorithm can be used in practical application.更多还原