基于SRUF特征点匹配的图像融合

【实例简介】基于SURF特征点匹配进行图像拼接

【实例截图】

【核心代码】function newImage = SURFAlign(file1, file2)
I1=imread(file1);%读取图片
I2=imread(file2);

%两幅图宽高可能不一致，为了方便显示，将窄的扩充
[w1,h1,~]=size(I1);
[w2,h2,~]=size(I2);
w=max(w1,w2);
h=max(h1,h2);

I1show=zeros(w,h,3);
I2show=zeros(w,h,3);
I1show(1:w1,1:h1, :)=I1;
I2show(1:w2,1:h2, :)=I2;

 figure;
 imshow(uint8([I1show,I2show]));%并排显示两幅待拼接图像
 title('待拼接图像');

img1=rgb2gray(I1); %将rgb转成灰度图，方便处理
img2=rgb2gray(I2);
imageSize=size(img1);
 
p1=detectSURFFeatures(img1);
p2=detectSURFFeatures(img2);%检测SURF特征点

[img1Features, p1] = extractFeatures(img1, p1);%使用64维向量表示特征描述子,
%第一个返回的参数即为每个特征点对应的特征描述子，第二个参数是特征点
[img2Features, p2] = extractFeatures(img2, p2);
boxPairs = matchFeatures(img1Features, img2Features);%特征描述子匹配
 
matchedimg1Points = p1(boxPairs(:, 1));%第二个参数:可以不加，因为其为n行1列的结构体数组
matchedimg2Points = p2(boxPairs(:, 2));

figure;
showMatchedFeatures(I1, I2, matchedimg1Points, matchedimg2Points, 'montage');
title('匹配的点 (包括内点和外点)');

[tform, inlierimg2Points, inlierimg1Points] = ...
estimateGeometricTransform(matchedimg2Points, matchedimg1Points, 'affine');%射影变换，tfrom映射点对1内点到点对2内点

figure;
showMatchedFeatures(I1, I2, inlierimg1Points, inlierimg2Points, 'montage');
title('匹配的点 (只包括内点)');


Rfixed = imref2d(size(I1));
[registered2, Rregistered] = imwarp(I2, tform);
%[registered1, Rregistered1] = imwarp(I1, tform);
figure;
imshowpair(I1,Rfixed,registered2,Rregistered,'blend');
title('直接融合');


[xlim, ylim] = outputLimits(tform, [1 imageSize(2)], [1 imageSize(1)]);%输出坐标范围 x:23.8~4334 y:-1844~1447
% 找到输出空间限制的最大最小值
xMin = min([1; xlim(:)]);%1
xMax = max([imageSize(2); xlim(:)]);%4334
 
yMin = min([1; ylim(:)]);%-1844
yMax = max([imageSize(1); ylim(:)]);%3000
 
% 全景图的宽高
width  = round(xMax - xMin);
height = round(yMax - yMin);
 
 
 
%创建2D空间参考对象定义全景图尺寸
xLimits = [xMin xMax];
yLimits = [yMin yMax];
panoramaView = imref2d([height width ], xLimits, yLimits);
 
% 变换图片到全景图.
unwarpedImage = imwarp(I1,projective2d(eye(3)), 'OutputView', panoramaView);
warpedImage = imwarp(I2, tform, 'OutputView', panoramaView);

newImage=unwarpedImage;
newImage=double(newImage);
 
balck1=(warpedImage(:,:,1)==0 & warpedImage(:,:,2)==0 & warpedImage(:,:,3)==0);
balck2=(newImage(:,:,1)==0 & newImage(:,:,2)==0 & newImage(:,:,3)==0);
black=and(balck1,balck2);
black=~black;
 
maskA = (warpedImage(:,:,1)>0 |warpedImage(:,:,2)>0 | warpedImage(:,:,3)>0);%变换图像掩膜
mask1 = (newImage(:,:,1)>0 | newImage(:,:,2)>0 | newImage(:,:,3)>0);%非变换图像掩膜
mask1 = and(maskA, mask1);%重叠区掩膜

[row,col] = find(mask1==1);
left = min(col);
right = max(col);%获得重叠区左右范围
up=min(row);
down=max(row);
mask = ones(size(mask1));

mask(:,left:right) = repmat(linspace(0,1,right-left 1),size(mask,1),1);%复制平铺矩阵
warpedImage=double(warpedImage);
warpedImage(:,:,1) = warpedImage(:,:,1).*mask;
warpedImage(:,:,2) = warpedImage(:,:,2).*mask;
warpedImage(:,:,3) = warpedImage(:,:,3).*mask;
 
mask(:,left:right) = repmat(linspace(1,0,right-left 1),size(mask,1),1);
newImage(:,:,1) = newImage(:,:,1).*mask;
newImage(:,:,2) = newImage(:,:,2).*mask;
newImage(:,:,3) = newImage(:,:,3).*mask;
 
newImage(:,:,1) = warpedImage(:,:,1) newImage(:,:,1);  %重构输出图像
newImage(:,:,2) = warpedImage(:,:,2) newImage(:,:,2);
newImage(:,:,3) = warpedImage(:,:,3) newImage(:,:,3);

newImage=uint8(newImage);
figure;
imshow(newImage);
title('渐入渐出融合');
end