SIFT contrast problem [closed]

asked 2015-07-14 06:38:11 -0600

neouyghur gravatar image

updated 2015-07-15 06:16:13 -0600

SIFT can't find corresponding key-points in two images whose colors are reversed. Could anyone give me any suggestion for handling this problem. Thank you. For more clear understanding, check this two images below:

Sift results: image description image description

Results should be: image description image description

edit retag flag offensive reopen merge delete

Closed for the following reason the question is answered, right answer was accepted by sturkmen
close date 2020-11-29 07:32:15.100005

Comments

No It's not possible because SIFT means Scale-invariant feature transform it's not invariant when you change exposition. For example gradient values are opposites, histogram are in reverse order and so on. Pre processing is needed. May be you should find image boundarie between positive and negative this help you to define a mask use to Inverts every bit of an array using bitwise_not function. After you can use SIFT

LBerger gravatar imageLBerger ( 2015-07-15 07:38:16 -0600 )edit

Sift have 8 different gradient direction, therefore different contrast will return different descriptor. But when we use four different gradient direction. it should return same descriptor. Am I right?

neouyghur gravatar imageneouyghur ( 2015-07-15 16:12:34 -0600 )edit

Gradients will have opposite sign

LorenaGdL gravatar imageLorenaGdL ( 2015-07-16 04:45:57 -0600 )edit
add a comment