Algorithm suggestion for texture image matching
Hello, I am trying to match texture images. what i tried,
- Feature detectors like sift, surf, orb,akage. (they could not found feature points for some images of my image DB)
- LBP (It kind of gives me matching image. But it seems like the matching is based on structure rather than color. say, the structure of two images are ok but the color is totally different.)
- Dominant Color + LBP (to remove the above color problem, 1st i found dominant color of image. then i tried LBP compare. i average two results. Now it is better. I also tried like taking 30% of 1st stage result and 70% of 2nd stage result and making the final result too.)
- Color histogram using Calchist + LBP ( i also tried to combine color hist compare result and LBP compare result)
So far, i have come to this point. Few times, results are not as expected, sometimes it is okey.
I would love to hear if anyone suggest any idea to do it more accurately.
Some sample images are like this:
Could be useful for an overview: From BoW to CNN: Two Decades of Texture Representation for Texture Classification.
Thank you so much. i will go through it! :)
is there a script suggestion gor this topic ??
The article suggested by Eduardo is a great resource.
Concerning the implementations, the Gabor filters can be a good start, as they are already in OpenCV.
Then, add some ML classifiers (like SVM, k-means, closest neighbor, bayesian estimator - also part of OpenCV) to discriminate the textures.
see also https://www.pyimagesearch.com/2015/12...