[Image Alignment] Align two images taken under different zooming and lighting conditions

I have 2 images taken the same object but under different zooming and lighting conditions (img1 and img2). I would like to apply a transformation matrix to img2, so that the result will be aligned with img1.

img1 (left) / img2 (right)

• If I place 2 images on top of each other, I got image on the left which shows 2 images are not aligned.
• If I apply a transform matrix A = [0.52 0 233; 0 1 0; 0 0 1] to img2, then combine, I got image on the right which shows kind of alignment.

I am trying to figure out how to find that transform matrix A. No need to be exact, close is good enough. I found A by manually try and error.

There are intensity-based image alignment methods I think cannot apply here because of the obvious difference in intensity.

Feature-based image alignment methods I tried included FAST, MinEigen, Harris, SIFT, SURF, KAZE, BRISK, ORB (link here)

https://www.mathworks.com/help/vision/ug/local-feature-detection-and-extraction.html

, but so far nothing is fruitful.

Because the y-axis is fixed, the only thing needed to adapt is the x-axis (scaleX and translateX).

Any ideas would be appreciated.

Thanks.

Answer 1

[Cris LaPierre]

I used the Registration Estimator App, which produced the following resutls.

img1 = imread('img1.jpg');
img2 = imread('img2.jpg');
[MOVINGREG] = registerImages(img2,img1)

MOVINGREG = struct with fields:
     FixedMatchedFeatures: [45x1 SURFPoints]
    MovingMatchedFeatures: [45x1 SURFPoints]
           Transformation: [1x1 affine2d]
          RegisteredImage: [512x512x3 uint8]
            SpatialRefObj: [1x1 imref2d]

imshowpair(img1,MOVINGREG.RegisteredImage)

function [MOVINGREG] = registerImages(MOVING,FIXED)
%registerImages  Register grayscale images using auto-generated code from Registration Estimator app.
%  [MOVINGREG] = registerImages(MOVING,FIXED) Register grayscale images
%  MOVING and FIXED using auto-generated code from the Registration
%  Estimator app. The values for all registration parameters were set
%  interactively in the app and result in the registered image stored in the
%  structure array MOVINGREG.

% Auto-generated by registrationEstimator app on 09-May-2025
%-----------------------------------------------------------


% Feature-based techniques require license to Computer Vision Toolbox
checkLicense()

% Convert RGB images to grayscale
FIXED = im2gray(FIXED);
MOVINGRGB = MOVING;
MOVING = im2gray(MOVING);

% Default spatial referencing objects
fixedRefObj = imref2d(size(FIXED));
movingRefObj = imref2d(size(MOVING));

% Detect SURF features
fixedPoints = detectSURFFeatures(FIXED,'MetricThreshold',649.525070,'NumOctaves',3,'NumScaleLevels',5);
movingPoints = detectSURFFeatures(MOVING,'MetricThreshold',649.525070,'NumOctaves',3,'NumScaleLevels',5);

% Extract features
[fixedFeatures,fixedValidPoints] = extractFeatures(FIXED,fixedPoints,'Upright',true);
[movingFeatures,movingValidPoints] = extractFeatures(MOVING,movingPoints,'Upright',true);

% Match features
indexPairs = matchFeatures(fixedFeatures,movingFeatures,'MatchThreshold',68.476562,'MaxRatio',0.684766);
fixedMatchedPoints = fixedValidPoints(indexPairs(:,1));
movingMatchedPoints = movingValidPoints(indexPairs(:,2));
MOVINGREG.FixedMatchedFeatures = fixedMatchedPoints;
MOVINGREG.MovingMatchedFeatures = movingMatchedPoints;

% Apply transformation - Results may not be identical between runs because of the randomized nature of the algorithm
tform = estimateGeometricTransform2D(movingMatchedPoints,fixedMatchedPoints,'affine');
MOVINGREG.Transformation = tform;
MOVINGREG.RegisteredImage = imwarp(MOVINGRGB, movingRefObj, tform, 'OutputView', fixedRefObj, 'SmoothEdges', true);

% Store spatial referencing object
MOVINGREG.SpatialRefObj = fixedRefObj;

end

function checkLicense()

% Check for license to Computer Vision Toolbox
CVTStatus = license('test','Video_and_Image_Blockset');
if ~CVTStatus
    error(message('images:imageRegistration:CVTRequired'));
end

end

Here's a screenshot of the app interface so you can see the approximate settings I used.

[TP]

This is great. Thanks!