search

hangong / ch

Demostration of color homography theorem with experiments in color correction, color feature matching, and color indexing, published in TPAMI 2017.
MIT License
7 stars 1 forks source link

Example usage #1

Open nikky4D opened 5 years ago

nikky4D commented 5 years ago

Thank you for posting this code. Do you have an example usage I can use as a test? I am working on low light enhancement and would like to test if your work can be used for low light image contrast enhancement

hangong commented 5 years ago

Hi. I'm not sure what is your expectation (input ? output? ). I guess this could be useful?

nikky4D commented 5 years ago

The inputs would be an image taken in night/low light condition (imgA), and an example daylight/bright light image (imgB). The expected output would be imgA with the lighting/illumination of imgB. The images, imgA and imgB, are of similar scenes. Does this make sense?

Thank you for the link

hangong commented 5 years ago

I'm afraid there is not an example for your case. I think you will need to tweak. A 2-D homograph match only gives you a 3x3 chromaticity correction matrix given the per-pixel RGB correspondences. The tonal RGB scaling can be done separately.

I guess, for your problem, you could try using a quick and dirty colour alignment method (e.g. histogram matching) which gives you a contrast transferred target image (i.e. per-pixel correspondences) maybe done in a lower resolution. Then, you enforce a 2D or 3D colour homography colour transfer approximation to get a artefact-reduced colour homography colour mapping (that link).

alt text