CODEWITHSUNDEEP
pythonopencvhomography
Saravanabalagi Ramachandran
Saravanabalagi Ramachandran

Reputation: 8889

Python: OpenCV findHomography inputs

I am trying to find the homography matrix for two images rgb and rotated using opencv in Python:

print(rgb.shape, rotated.shape)
H = cv2.findHomography(rgb, rotated)
print(H)

And the error I get is

(1080, 1920, 3) (1080, 1920, 3)
---------------------------------------------------------------------------
error                                     Traceback (most recent call last)
<ipython-input-37-26874dc47f1f> in <module>()
      1 print(rgb.shape, rotated.shape)
----> 2 H = cv2.findHomography(rgb, rotated)
      3 print(H)

error: OpenCV(3.4.1) C:\projects\opencv-python\opencv\modules\calib3d\src\fundam.cpp:372: error: (-5) The input arrays should be 2D or 3D point sets in function cv::findHomography

I also tried with cv2.findHomography(rgb[:,:,0], rotated[:,:,0]) to see if the channels or channel ordering is causing any problem, but it's not working for even 2D matrix.

How should the input be?

Upvotes: 3

Views: 11692

Answers (1)

Saravanabalagi Ramachandran
Saravanabalagi Ramachandran

Reputation: 8889

cv2.findHomography() doesn't take in two images and return H.

If you need to find H for two RGB images as np.arrays:

import numpy as np
import cv2

def findHomography(img1, img2):

    # define constants
    MIN_MATCH_COUNT = 10
    MIN_DIST_THRESHOLD = 0.7
    RANSAC_REPROJ_THRESHOLD = 5.0

    # Initiate SIFT detector
    sift = cv2.xfeatures2d.SIFT_create()

    # find the keypoints and descriptors with SIFT
    kp1, des1 = sift.detectAndCompute(img1, None)
    kp2, des2 = sift.detectAndCompute(img2, None)

    # find matches
    FLANN_INDEX_KDTREE = 1
    index_params = dict(algorithm=FLANN_INDEX_KDTREE, trees=5)
    search_params = dict(checks=50)

    flann = cv2.FlannBasedMatcher(index_params, search_params)
    matches = flann.knnMatch(des1, des2, k=2)

    # store all the good matches as per Lowe's ratio test.
    good = []
    for m, n in matches:
        if m.distance < MIN_DIST_THRESHOLD * n.distance:
            good.append(m)


    if len(good) > MIN_MATCH_COUNT:
        src_pts = np.float32([kp1[m.queryIdx].pt for m in good]).reshape(-1, 1, 2)
        dst_pts = np.float32([kp2[m.trainIdx].pt for m in good]).reshape(-1, 1, 2)

        H, _ = cv2.findHomography(src_pts, dst_pts, cv2.RANSAC, RANSAC_REPROJ_THRESHOLD)
        return H

    else: raise Exception("Not enough matches are found - {}/{}".format(len(good), MIN_MATCH_COUNT))

Note:

  • Tested on Python 3 and OpenCV 3.4
  • You need opencv-contrib-python package because SIFT has patent issues and has been removed from opencv-python
  • This gives the H matrix for transforming img1 and overlap it on img2. If you're wondering how to do this, it's here

Upvotes: 3

Related Questions