There is a well known process -- flat fielding -- that is used to correct for uneven illumination in images in scientific and astronomical images. See here:
http://imagej.net/Image_Intensity_Proce ... correction
I'm wondering if anyone has tried this method to eliminate or reduce illumination gradients that many lighting configurations? I'm sure it is a very compute intensive process but is something that could be useful when it's important to get evenly lit images and lighting is not optimal.
Flat fields to correct for uneven illumination -- has anyone tried this?
Moderator: peterZ
Re: Flat fields to correct for uneven illumination -- has anyone tried this?
You could try CLAHE (Contrast Limited Adaptive Histogram Equalization). With OpenCV it is 6 lines of python: https://opencv-python-tutroals.readthed ... ation.html
- daniel_reetz
- Posts: 2812
- Joined: 03 Jun 2009, 13:56
- E-book readers owned: Used to have a PRS-500
- Number of books owned: 600
- Country: United States
- Contact:
Re: Flat fields to correct for uneven illumination -- has anyone tried this?
Yeah, in the very first scanner, Aaron Clarke and I would take a picture of a blank piece of paper, and then subtract the inverse of that from the page image. It works, as long as your cameras have fixed settings and the camera-platen relationship doesn't change. If it changes at all, you need a new reference image. If you want to try this, it's easy to do in Photoshop with just two layers.
If you're only interested in correcting for vignetting or low-frequency falloff, you can also blur an image of a blank piece of paper and subtract that.
Here are some other weird ideas around lighting: http://www.danreetz.com/blog/2011/01/05 ... -scanning/
If you're only interested in correcting for vignetting or low-frequency falloff, you can also blur an image of a blank piece of paper and subtract that.
Here are some other weird ideas around lighting: http://www.danreetz.com/blog/2011/01/05 ... -scanning/
Re: Flat fields to correct for uneven illumination -- has anyone tried this?
We did this for years in Capture One software. It's called LCC. Works really well.