基于位移旋转不变特征描述符sRIFD实现图像配准附matlab代码论文复现

最新推荐文章于 2024-04-25 21:38:01 发布

天天Matlab代码科研顾问

最新推荐文章于 2024-04-25 21:38:01 发布

阅读量53

点赞数

分类专栏：图像处理文章标签： matlab 开发语言

本文链接：https://blog.csdn.net/m0_57702748/article/details/134452324

版权

图像处理专栏收录该内容

369 篇文章 28 订阅

订阅专栏

✅作者简介：热爱科研的Matlab仿真开发者，修心和技术同步精进，

代码获取、论文复现及科研仿真合作可私信。

🍎个人主页：Matlab科研工作室

🍊个人信条：格物致知。

更多Matlab完整代码及仿真定制内容点击👇

智能优化算法神经网络预测雷达通信无线传感器电力系统

信号处理图像处理路径规划元胞自动机无人机

🔥 内容介绍

特征描述符（例如尺度不变特征变换（SIFT））已经广泛应用于各种计算机视觉应用程序，并且已经提出了许多方法来使描述符适应多传感器图像。最近，辐射变化不敏感特征变换（RIFT）利用了相位一致性，该相位一致性对于光照和辐射变化是不变的，用于构建多传感器图像的描述符。然而，即使具有准确的关键点主方向，RIFT在某些旋转度的图像上表现不佳。为了解决这个问题，本文设计了一个SoMIM（二阶最大指数图），使用Log-Gabor滤波器的方向响应，然后构建了一个新颖的描述符sRIFD（位移旋转不变特征描述符），而无需计算主方向。sRIFD在几乎所有旋转角度的图像上表现均匀良好。为了评估sRIFD的性能，我们使用了三个图像数据集，红外-光学、热-光学和光学-光学数据集。定量结果表明，在不同旋转角度下，sRIFD在多传感器图像数据集上表现更好。

📣 部分代码

% This is a samplest implementation of the proposed sRIFD algorithm. In this implementation,...% rotation invariance part and corner point detection are not included.function [des_m1,des_m2,or1,or2] = sRIFD_rotation_invariance(im1,im2,s,o)% m1 and m2 are the maximum moment maps;% eo1{s,o} = convolution result for scale s and orientation o.% The real part is the result of convolving with the even symmetric filter,% the imaginary part is the result from convolution with the odd symmetric filter.[m1,M1,or1,~,~,eo1,~] = phasecong3(im1,s,o,3,'mult',1.6,'sigmaOnf',0.75,'g', 3, 'k',1);[m2,M2,or2,~,~,eo2,~] = phasecong3(im2,s,o,3,'mult',1.6,'sigmaOnf',0.75,'g', 3, 'k',1);a=max(m1(:)); b=min(m1(:)); m1=(m1-b)/(a-b);a=max(m2(:)); b=min(m2(:)); m2=(m2-b)/(a-b);a=max(M1(:)); b=min(M1(:)); M1=(M1-b)/(a-b);a=max(M2(:)); b=min(M2(:)); M2=(M2-b)/(a-b);% FAST detector on the maximum moment maps to extract edge feature points.m1_points = detectFASTFeatures(m1,'MinContrast',0.0001,'MinQuality',0.0001);m2_points = detectFASTFeatures(m2,'MinContrast',0.0001,'MinQuality',0.0001);m1_points=m1_points.selectStrongest(5000);   %number of keypoints can be set by usersm2_points=m2_points.selectStrongest(5000);M1_points = detectFASTFeatures(M1,'MinContrast',0.0001,'MinQuality',0.0001);M2_points = detectFASTFeatures(M2,'MinContrast',0.0001,'MinQuality',0.0001);M1_points=M1_points.selectStrongest(1);   %number of keypoints can be set by usersM2_points=M2_points.selectStrongest(1);% RIFT descriptorim11=imrotate(im1,15);im22=imrotate(im2,15);[m1,~,or1,~,~,eo11,~] = phasecong3(im11,s,o,3,'mult',1.6,'sigmaOnf',0.75,'g', 3, 'k',1);[m2,~,or2,~,~,eo22,~] = phasecong3(im22,s,o,3,'mult',1.6,'sigmaOnf',0.75,'g', 3, 'k',1);pt1=[m1_points.Location;M1_points.Location];pt2=[m2_points.Location;M2_points.Location];des_m1 = fct_calc_log_sRIFD_no_dominant(im1, pt1,eo1,  s,o);des_m2 = fct_calc_log_sRIFD_no_dominant(im2, pt2,eo2, s,o);