基于自适应分数阶微分的引导滤波及其应用

发布时间：2018-01-07 02:34

本文关键词：基于自适应分数阶微分的引导滤波及其应用　出处：《计算机应用研究》2017年01期 　论文类型：期刊论文

【摘要】：引导滤波算法具有保边平滑的功能,但传统引导滤波方法容易导致图像平滑区域过度模糊、细节丢失的问题。为了使引导滤波在保持高频信息的同时结构化输出低频灰度,提出了一种基于自适应分数阶微分的引导滤波算法。以分数阶微分理论为基础定义了分数阶微分掩模,并结合图像梯度、二维信息熵和局部方差权值构造了自适应分数阶微分阶数函数来有效检测图像纹理和梯度变化,从而将图像局部特性转移到引导图像中,确保在平滑去噪的同时保持图像纹理细节。实验结果表明,算法具有良好的边缘和纹理保持特性,将算法运用到基于PCA和SVM的人脸识别图像预处理中,能一定程度地提升人脸识别率。
[Abstract]:Guide the filtering algorithm with edge preserving smoothing function, but the traditional bootstrap filtering method is easy to cause image smoothing over regional fuzzy problems. In order to make the details of the missing guide filter in keeping the high frequency information and structured output low frequency gray level, put forward a kind of adaptive filtering algorithm guide based on fractional differential. The fractional differential theory the definition of fractional differential mask, and combining with the image gradient, two-dimensional entropy and local variance weights to construct adaptive fractional order differential function to detect the image texture and gradient changes, which will be transferred to the local characteristics of image guided image, at the same time to ensure smooth image denoising in texture detail. The experimental results show that the and the algorithm has good edge and texture preserving properties, the algorithm is applied to the preprocessing of face image recognition based on PCA and SVM, can to some extent. Improve face recognition rate.

【作者单位】：华中师范大学计算机学院;
【基金】：国家自然科学基金资助项目(61501198,41101425,41201413) 湖北省自然科学基金面上资助项目(2014CFB461) 武汉市青年科技晨光计划资助项目(2014072704011248) 华中师范大学中央高校基本科研业务费资助项目(CCNU14A05017)
【分类号】：TP391.41;TN713
【正文快照】： 0引言在计算机视觉和图像处理领域的许多应用中,保边平滑技术已经被广泛地采用。为了在滤除噪声的同时保持边缘,To-masi等人[1]提出了双边滤波(bilateral filter)算法,它是一种结合图像的空间邻近度和像素值相似度的非线性滤波方法,具有保边去噪的特性。He等人[2]在此基础上提

【相似文献】