skimage.feature.match_descriptors() should return matching scores

[chuong]

Matching scores (or distance between descriptors) are needed to determine which matches are reliable. Say, I want to get 10 best matches from two sets of features. This is only possible when matching scores are available.

Matching scores can be return as a second output of skimage.feature.match_descriptors()

[ajaybhat]

As I understand it, the 'score' simply means the difference between the index of a matched descriptor pair. So for the match [[0, 2], [1, 1], [3, 8]] the scores would be [2, 0, 5] is that right?

[ahojnnes]

That depends on the distance metric you use..., also you need to sets of features to compute the distances. Where is the second set?

----- On May 27, 2015, at 11:39 AM, Ajay Bhat notifications@github.com wrote:

As I understand it, the 'score' simply means the difference between the index of
a matched descriptor pair. So for the match [[0, 2], [1, 1], [3,
8] the scores would be [2, 0, 5] is that right?

---

Reply to this email directly or view it on GitHub:
#1297 (comment)

[ajaybhat]

np.array([[1, 2], [3, 4], [4, 5], [6, 2], [1, 2], [7, 2], [1, 2], [3, 4], [2, 2]])

np.array([[4, 2], [3, 1], [1, 2], [4, 2], [4, 2], [3, 2], [0, 2], [1, 2], [6, 2]])

This is the features being used which gives the match [[0, 2], [1, 1], [3, 8]]. Scores for metric None should be [2, 0, 5].
What will it be for other metrics?

[kkmou]

Any solution to this issue so far?