中文摘要：

字符矫正是光学字符识别（OCR）系统预处理过程中的重要步骤,针对传统的增广拉格朗日乘子法（ALM）求解字符矫正问题时收敛性和计算速度的不足,本文研究了并行分离的增广拉格朗日乘子法,综合考虑字符矫正模型的建立过程,提出并行分离方法与ALM相结合的思想解决字符矫正问题。用并行方式将迭代问题分解成3个子问题,计算时能够同时求解分解后的这3个子问题,然后进行凸组合,最后收敛到问题的最优解。实验结果表明,本文算法能够快速准确地对变形的字符图像进行矫正,并且具有良好的实时性和适应性,可用于OCR系统的矫正预处理中,提高OCR系统的识别率。

英文摘要：

Character modification is an important step in the preprocessing of optical character recognition （OCR） system, which determines the OCR performance. Classic augmented Lagrangian multiplier （ALM） method is still insufficient to solve low rank character image. For the lack of convergence and computation rate of traditional augmented Lagrangian multiplier method,this paper proposes a new par allel splitting augmented Lagrangian multiplier method （PSALM）, which combines the parallel separation method with the ALM algorithm. The optimal solution of the original problem can be attained using three iterated sub problems. Experimental results show that parallel splitting ALM can correct the dis torted character image efficiently and accurately and has sufficient adaptability and real-time ability, which can be used for the preprocessing of OCR system to improve OCR performance.