基于信息特征子空间均匀遍历的图像去噪算法  

Image Denoising Algorithm Based on Information Feature Subspace Uniform Traversal

作　　者：陆钊[1,2] 李超建[2] LU Zhao LI Chao-jian(Guangxi Universities Key Lab of Complex System Optimization and Big Data Processing,Yulin 573000, China School of Computer Science and Engineering, Yulin Normal University, Yulin 573000, China)

机构地区：[1]广西高校复杂系统优化与大数据处理重点实验室,广西玉林537000 [2]玉林师范学院计算机科学与工程学院,广西玉林537000

出　　处：《微电子学与计算机》2017年第3期105-109,共5页Microelectronics & Computer

基　　金：广西高校科学技术研究项目(KY2015YB241;KY2015LX300;2013LX112);玉林师范学院创新创业教育课程建设项目(2016CXCYKC15)

摘　　要：在强干扰环境下采集的远程图像通常含有大量的噪声,成像质量不好,需要进行图像去噪处理,提高图像识别和成像性能.传统的图像去噪方法采用小波变换噪点平滑处理算法,在信噪比比较小的情况下出现视觉偏移,图像去噪效果不好.对此,提出一种基于信息特征子空间均匀遍历的图像去噪算法.在三角网纹理分区结构模型的基础上,通过相互近似正交的直线模式建立含噪图像直线模式的方向关系,进行图像滤波设计,实现噪点分离,采用全局非显著性突变信息提取方法,基于均遍历寻址方式,实现对图像噪点的特征遍历滤除,实现图像去噪算法改进.仿真结果表明,该算法具有较好的图像去噪性能,提高对含噪模糊图像的成像质量和识别概率,峰值信噪比较高,在远程模糊图像识别等领域具有应用价值.Remote image acquisition in strong interference environment usually contain a lot of noise, the image quality is not good, the need for image denoising, image recognition and imaging performance improvement. The traditional image denoising methods using wavelet transform noise smoothing algorithm, the emergence of visual offset ratio is small in the case of signal to noise, image denoising effect is not good. This paper proposed an image de-noising algorithm information feature subspace based on uniform traversal. Based on triangular mesh texture partition structure model, the establishment of the noisy image linear model of direction relations by linear mode between approximately orthogonal design implementation, image filtering, noise isolation, with global non obvious mutation information extraction methods, they are ergodic addressing way based on the characteristics of image noise filtering to realize the traverse, implementation improved image denoising algorithm. The simulation results show that the algorithm has better performance in image denoising, it can improve on the quality of imaging and identification probability fuzzy image with noise, peak signal to noise is relatively high, and it has application value in the fields of remote fuzzy image recognition.

关 键 词：图像 去噪 特征子空间 均匀遍历 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]