2
$\begingroup$

I already put this question in stackoverflow.com. But still didn't get an answer. I want to compare similarity between below images. Acording to my requirements I want to identify all of these images as similar, since it has use the same color, same clip art. The only difference in these images are rotation ,scale and the placement of the clip art. Since all 3 t-shirts has used the same color and clip art I want to identify all 3 images as similar. I tried out the method described in hackerfactor.com. But it doesn't give me correct result acording to my requirements. How to identify all these images as similar? Thanks in advance.

enter image description here enter image description here enter image description here

Improve this question
$\endgroup$
2
  • $\begingroup$ If to speak about general statistical techniques rather than specialized image-analytical algorithms then procrustes analysis can show that the three clipart objects (extracted from the t-shirts) are identical; can compare them, in general. The main problem with procrustes is that it needs to know beforehand which point on one object corresponds to the given point of another object. For example, when you digitize (invent X-Y coordinates) of the three faces, the beak end "landmark" must be point i in all the three datasets; $\endgroup$
    – ttnphns
    Commented Nov 1, 2015 at 9:26
  • $\begingroup$ (cont.) the cirrus endpoint must be point j in all the three datasets, and so on. So, the problem of finding or settint the correspondence of the "same" points must be already solved somehow, this is the main challenge. It could be solved by various approaches, including those engaged in computation of a distance matrix between points within each face. $\endgroup$
    – ttnphns
    Commented Nov 1, 2015 at 9:30

2 Answers 2

Reset to default
2
$\begingroup$

Try the SIFT or SURF features. They are scale, rotation and translation invariant. Run k-means on the descriptors and go from feature vectors to integer codes (cluster centers). Then build Bag of Words representation per image, basically counting each cluster codes. It will be beneficial to apply TF-IDF normalization on those counts. The three images should translate into some global representation which among them should be very similar (e.g. cosine similarity).

Improve this answer
$\endgroup$
2
  • $\begingroup$ Is there any libraries for php to achieve SIFT/SURF features? If so, can you provide me a link? $\endgroup$
    – user3894236
    Commented Jul 9, 2015 at 6:01
  • $\begingroup$ PHPimage seems to have some abilities but I would recommend you check out Python modules for the matter... CV people code in Python routinely and you will have far more options to choose from. $\endgroup$
    – usεr11852
    Commented Nov 1, 2015 at 8:40
0
$\begingroup$

Try to use feature fusion technique by fusing colour, texture, and shape features using one of standard feature fusion methods (Early feature fusion or Late feature fusion). Before any fusion process you should normalize the feature values to be in the same scale (0,1). Try also to read the paper in:

http://ajbasweb.com/old/ajbas/2015/106-117.pdf

Improve this answer
$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.