图像的HOG特征提取,并绘制直方图
%Image descriptor based on Histogram of Orientated Gradients for gray-level images. This code
%was developed for the work: O. Ludwig, D. Delgado, V. Goncalves, and U. Nunes, 'Trainable
%Classifier-Fusion Schemes: An Application To Pedestrian Detection,' In: 12th International IEEE
%Conference On Intelligent Transportation Systems, 2009, St. Louis, 2009. V. 1. P. 432-437. In
%case of publication with this code, please cite the paper above.
function H=HOG(Im)
nwin_x=3;%set here the number of HOG windows per bound box
nwin_y=3;
B=9;%set here the number of histogram bins
[L,C]=size(Im); % L num of lines ; C num of columns
H=zeros(nwin_x*nwin_y*B,1); % column vector with zeros
m=sqrt(L/2);
if C==1 % if num of columns==1
Im=im_recover(Im,m,2*m);%verify the size of image, e.g. 25x50
L=2*m;
C=m;
end
Im=double(Im);
step_x=floor(C/(nwin_x+1));
step_y=floor(L/(nwin_y+1));
cont=0;
hx = [-1,0,1];
hy = -hx';
grad_xr = imfilter(double(Im),hx);
grad_yu = imfilter(double(Im),hy);
angles=atan2(grad_yu,grad_xr);
magnit=((grad_yu.^2)+(grad_xr.^2)).^.5;
for n=0:nwin_y-1
for m=0:nwin_x-1
cont=cont+1;
angles2=angles(n*step_y+1:(n+2)*step_y,m*step_x+1:(m+2)*step_x);
magnit2=magnit(n*step_y+1:(n+2)*step_y,m*step_x+1:(m+2)*step_x);
v_angles=angles2(:);
v_magnit=magnit2(:);
K=max(size(v_angles));
%assembling the histogram with 9 bins (range of 20 degrees per bin)
bin=0;
H2=zeros(B,1);
for ang_lim=-pi+2*pi/B:2*pi/B:pi
bin=bin+1;
for k=1:K
if v_angles(k)<ang_lim
v_angles(k)=100;
H2(bin)=H2(bin)+v_magnit(k);
end
end
end
H2=H2/(norm(H2)+0.01);
H((cont-1)*B+1:cont*B,1)=H2;
end
end
D106
相关文章
- Word控件Spire.Doc 【图像形状】教程(9): 如何在 C# 和 VB.NET 中从 word 文档中删除形状
- Word控件Spire.Doc 【图像形状】教程(8): 如何借助C#/VB.NET在 Word 中插入艺术字
- Flutter图像绘制原理深入分析
- Google Earth Engine——NOAA/GOES/16/和17/MCMIPF地球静止气象卫星云层和水分图像产品的(分辨率都是2公里)
- 【形态学检测】基于形态学处理的遥感图像的目标提取
- 图像的QPSK无线传输仿真
- 利用OpenCV的函数calcHist()计算出图像的直方图数据后绘制图像的直方图
- 灰度图像的对数变换原理及OpenCV代码实现
- 我对图像金字塔的理解及OpenCV下的实现代码
- python数字图像处理(3):图像像素的访问与裁剪
- 《响应式Web图形设计》一13.1 缩放图像和媒体
- openCV 图像拼接
- 在线绘制2维/3维函数图像、求导网站
- 【图像处理】——图像灰度直方图的绘制(直接调用函数和自定义函数)
- 帮助识别卫星图像,美国情报部门用人工智能当间谍
- 图像相似性度量—— 归一化互信息实现