late fusion over color and texture descriptors

asked 2019-12-17 00:58:45 -0600

arafat gravatar image

Any idea on how to do late fusion over color and texture descriptors?

I am doing a texture matching project. I want to merge color descriptor (I used calchist()) and texture descriptor (i used LBP). From a paper, i get to know that late fusion is a way to do it. They meant sum or product of sum to do it but unfortunately i did not get it.

Now i am comparing the histogram of color and LBP among test image and Database images and computing the average of them.

Any idea on how to do this late fusion on Color descriptor and texture descriptors?

Thanks!

edit retag flag offensive close merge delete

Comments

"late fusion" -- what is that ? please explain

berak gravatar imageberak ( 2019-12-17 01:29:52 -0600 )edit

Its one of the multimodal fusion techniques. I found it from this

arafat gravatar imagearafat ( 2019-12-17 01:43:31 -0600 )edit

what do you do with your LBP patterns ? how does your "descriptor" look like ? how much data do you have ?

berak gravatar imageberak ( 2019-12-17 03:30:17 -0600 )edit
1

for details about what i am doing is here

I used LBP as texture descriptor. I followed this

I stored database image's LBP histogram and later compared with input image's LBP by CompareHist() of opencv. Having the lowest value gives the most similar match. But using LBP was not solving the color problem. So i tried Dominant color or Color histogram (using calchist()) 1st and then LBP.

In that late fusion, 1st stage's calchist() output and LBP output can be merged. I dnt know how.

Now i take the average of Color histogram's comparehist() value and LBP's histogram's comparehist().

I have 50+ types of textures of total 676 images.

arafat gravatar imagearafat ( 2019-12-17 03:41:24 -0600 )edit
add a comment