【作者】 邹建成； 车冬娟；

【Author】 ZOU Jian-cheng,CHE Dong-juan (Institute of Image Processing and Pattern Recognition,North China University of Technology,Beijing 100144,China)

【机构】 北方工业大学 图像处理与模式识别研究所；

【摘要】 图像在获取、传输等过程中会受到干扰,造成图像模糊不清。传统去模糊方法需要根据模糊原因估计出相应的点扩散函数(PSF),然后进行反卷积估计出原始图像。本文借鉴K-SVD算法,利用对模糊图像进行训练得到自适应字典;并利用计算量相对较小且收缩速度较快的PCD算法来求解稀疏解。实验结果表明,相对传统图像去模糊算法和固定字典算法,自适应的字典学习算法能达到更好的去模糊效果。更多还原

【Abstract】 Image can be blurred in the process of acquiring and transport.The traditional deblurring methods need to estimate the corresponding point spread function(PSF) according to the blurring reason, then estimate the original image by reversing convolution.In this paper,K-SVD algorithm will be used for the reference of requiring adaptive and trained dictionary;at the same time,PCD algorithm will be used for researching the sparse solution for its relatively small amount of calculation and fast shrinking.The experimental results show that,compared to the traditional image deblurring algorithms or fixed dictionary algorithms,adaptive dictionary learning algorithm can ger better result.更多还原