找到两幅图像重叠区域与MATLAB

Question

我想找到使用MATLAB中的两个图像的重叠区域。

Answer 1

如果您的最终目标是拼接全景，您可能需要考虑this code。在任何情况下，为了获得重叠区域，您需要首先注册图像（了解它们如何重叠 - 或者更加数学地说：找到从图像1到图像2的转换）。为此，您需要在两张图像中找到匹配点。

在下面的代码的确，对于两个图像（由this code启发，它使用较旧的MATLAB函数）。

%% >>>>>>> load images and calculate their transformation <<<<<<< %% 
im1 = imread('1.png'); 
im2 = imread('2.png'); 
imshowpair(im1, im2, 'montage'); 

% calculate features on grayscale image 
im1g = rgb2gray(im1); 
im2g = rgb2gray(im2); 
points1 = detectSURFFeatures(im1g); 
[features1, points1] = extractFeatures(im1g, points1); 
points2 = detectSURFFeatures(im2g); 
[features2, points2] = extractFeatures(im2g, points2); 

% Find correspondences between im1 and im2 
indexPairs = matchFeatures(features1, features2, 'Unique', true); 
matchedPoints1 = points1(indexPairs(:,1), :); 
matchedPoints2 = points2(indexPairs(:,2), :); 

% Identity transformation 
transform_eye = projective2d(eye(3)); 
% Estimate the transformation between im1 and im2 
% we use a 'similarity' transform (translation/rotation), which treats the 
% images as rigid bodys. 'affine'/'projective' transformations allow for 
% warping the images itself (the overlap might not be a rectangle). 
transform = estimateGeometricTransform(matchedPoints1, matchedPoints2,... 
    'similarity', 'Confidence', 99.9, 'MaxNumTrials', 2000); 

%% >>>>>>> apply transformation to images <<<<<<< %% 

% create a world coordinate system (RF) that has space to store 
% the reference image (im1) and the transformed image (im2) 
R2 = imref2d(size(im2)); 
[~, R2T]=imwarp(im2,R2,transform); 
xLimits=[min(0.5,R2T.XWorldLimits(1)) max(size(im1,2), R2T.XWorldLimits(2))]; 
yLimits=[min(0.5,R2T.YWorldLimits(1)) max(size(im1,1), R2T.YWorldLimits(2))]; 
width = round(xLimits(2) - xLimits(1)); 
height = round(yLimits(2) - yLimits(1)); 
RF = imref2d([height width], xLimits, yLimits); 

% transform both images with regard to the world coordinate system RF 
im1t=imwarp(im1,transform_eye,'OutputView',RF); % im1 stays in place (identity transform) 
im2t=imwarp(im2,transform,'OutputView',RF); % im2 is transformed 

% visualize result 
imOverlay = im1t/2 + im2t/2; 
imshow(imOverlay); 

%% >>>>>>> get the overlap area only <<<<<<< %% 
% if you only want the overlap area, apply the transform to image masks 
im1bw = ones(size(im1)); % mask1 
im2bw = ones(size(im2)); % mask2 
im1bwt=imwarp(im1bw,transform_eye,'OutputView',RF); % im1 stays in place (identity transform) 
im2bwt=imwarp(im2bw,transform,'OutputView',RF); % im2 is transformed 

% visualize result 
maskOverlap = im1bwt + im2bwt - 1; 
imshow(maskOverlap); 
% maskOverlap is a bw image that contains 'true' for overlap pixels 
% you can use that for cropping imOverlay or 
% use bwarea or regionprops to calculate the area