压缩感知的人脸图像去噪.pdfVIP

VoL 20 No.5 第20 卷第5 期 哈尔滨理工大学学报 JOURNAL OF HARBIN UNIVERSITY OF SCIENCE AND TECHNOLOGY Oct. 2015 2015 年 10 月 压缩感知的人脸图像去躁 林克正， 魏颖， 程卫月 (哈尔读理工大学计算机科学与技术学院，黑龙江哈尔滨 15∞80) 摘 要:为解决人脸识别领域的噪声图像恢复问题，提出一种压缩感知的人脸图像去嗓算法， 协同稀疏性度量(collaborative sparse measure , CSM). CSM 算法利用困像的先验知识，用一个域将 图像稀疏表示，将图像的二维稀疏表示和三维稀疏表示同时进行自适应混合空间域转换，利用增广 拉格朗日技术求解.实验结果表明，CSM 算法的信嗓比明显高于传统算法的信噪比，具有高效性. 关键词:压缩感知;协同稀疏性度量;圈像去噪;空间域 DOI: 10. 15938/j.jhus t. 2015. 05. 018 中图分类号: TP39 1. 4 文献标志码:A 文章编号: 1 ∞7- 2683 (2015 )05-∞91-06 Face Image Denoising of Compressed Sensing UN Ke-zheng , WEI Ying , CHENG Wei-yue (Schω1 of Computer Science and Technology , Harbin University of Science and Technology , Harbin 15∞80 ， China) Abstract: To solve the problem of noisy image restoration in face recognition area , a algorithm of face im- age de-noising of compressed sensing was proposed in this paper , called Collaborative Sparse Measure ( CSM) . The algorithm of CSM used a priori knowledge of image , image sparse representation with a domain , and the sparse representation of two-dimensional and three-dimensional simultaneously did adaptive hybrid spa- tial domain conversion , and solved problem with Augmented Lagrangian technique. The result of the experi- ment indicates that the signal to noise ratio of CSM algorithm is superior to that of the traditional algorithm , which has a high efficiency. Keywords :compression perception; collaborative sparse measurement; image de-noising; spatial domain (tree