如何使用 L2 范数（python）在描述符中查找匹配项？

Question

I need some clarification in order to understand how image matching work. Basically, I calculated keypoints (keypoints1, keypoints2) of img1 and img2. After I calculated descriptors (desc1, desc2). After that I used cdist function from scipy library to calculate the L2 norm of those descriptors. Now I do not understand how to find the matches using distances calculated. I don't understand how to find appropriate indexes that match. If possible can you show in python vector examples? (Like desc1 = [[1,2],[3,4]] and desc2 = [[5,6],[0.9, 2.1]]. Clearly here desc1[0] and desc2[1] has the minimum distance.)

Answer 1

Do you mean something along those lines?

scores=scipy.spatial.distance.cdist(desc1,desc2)

matches=scores.argsort(1)

In the snippet above, scores[i,j] contains the matching score between desc1[i] and desc2[j] , and

matches[i]=[index_1, ..., index_k] 

is such that desc2[index_1] is the closest match to desc1[i] , desc2[index_2] is the second best match to desc1[i] and so on.