结合离散正余弦变换的快速梯度矢量流算法  

Fast gradient vector flow algorithm combined with discrete sine and cosine transform

作　　者：许贤泽[1] 陈孟 徐逢秋[1] 

机构地区：[1]武汉大学电子信息学院

出　　处：《华中科技大学学报（自然科学版）》2018年第1期58-62,共5页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家重点基础研究发展计划资助项目(2011CB70794)

摘　　要：为了提高求解梯度矢量流（GVF）的效率和准确度,在狄利克雷或诺依曼边界条件下提出了非精确拉格朗日离散正弦梯度矢量流（IALM-DST-GVF）和非精确拉格朗日离散余弦梯度矢量流（IALM-DCT-GVF）快速算法.两种算法在非精确增广拉格朗日优化算法基础上,结合了离散正弦和余弦变换.其算法时间复杂度均为O（CNlgN）（其中C为迭代次数,N为图像像素数量）.在相同的环境下采用C＋＋语言编码验证,结果表明：提出的算法比当前主流GVF算法效率更高,并且边界上的GVF域比IALM-GVF算法准确.To improve the efficiency and accuracy of gradient vector flow（GVF）,inexact augment Lagrange method-discrete sine transform-gradient vector flow（IALM-DST-GVF） and inexact augment Lagrange method-discrete cosine transform-gradient vector flow（IALM-DCT-GVF） algorithms were proposed with the Dirichlet or Neumann boundary conditions. Based on the inexact augmented Lagrange optimization method,the algorithms combined discrete sine and cosine transform methods. Their time complexity were both O（CNlgN）,where C was the iteration times and N was the number of pixels. The proposal was verified in the same environment by C＋＋ programming.The result show that the proposed algorithms are more efficient than present GVF algorithms,and meanwhile the GVF on boundaries are more exact than the IALM-GVF algorithm.

关 键 词：梯度矢量流 离散正弦变换 离散余弦变换 增广拉格朗日方法 狄利克雷边界条件 诺依曼边界条件 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]