How can I use Homography?

Question

I am developing a program where I receive 2 pictures of the same scene, but one of them has a distortion:

Mat img_1 = imread(argv[1], 0);  // nORMAL pICTURE
Mat img_2 = imread(argv[2], 0);  // PICTURE WITH DISTORTION

AND I WOULD LIKE TO EVALUATE THE DISTORTIONS' PATTERN AND BE ABLE TO COMPENSATE IT

I AM ALREADY ABLE TO FIND THE KEYPOINTS AND I WOULD LIKE TO KNOW IF I CAN USE THE FUNCTION cv::findHomography for this... In any case, how to do so?

Answer 1

A homography will map one image plane to another. That means that if your distortion can be expressed as a 3x3 matrix, findHomography is what you want. If not, then it isn't what you want. It takes two vectors of corresponding points as input and will return the 3x3 matrix that best represents the transform between those points.

Answer 2

Alright, so suppose I've two pictures (A and B) slightly distorted one from the other, where there are translation, rotation and scale differences between them (for example, these pictures:)

---

Ssoooooooo what I need is to apply a kind of transformation in pic B so it compensates the distortion/translation/rotation that exists to make both pictures with the same size, orientation and with no translation

I've already extracted the points and found the Homography, as shown bellow. But I don'know how to use the Homography to transform Mat img_B so it looks like Mat img_A. Any idea?

//-- Localize the object from img_1 in img_2
std::vector<Point2f> obj;
std::vector<Point2f> scene;

for (unsigned int i = 0; i < good_matches.size(); i++) {
    //-- Get the keypoints from the good matches
    obj.push_back(keypoints_object[good_matches[i].queryIdx].pt);
    scene.push_back(keypoints_scene[good_matches[i].trainIdx].pt);
}

Mat H = findHomography(obj, scene, CV_RANSAC);

Cheers,