05-28-2023 05:17 AM
we are trying to remove haze from an hazy image with black channel prior. We split the hazy image to three different channels(R,G&B) and compared each pixels of different channels image and took the pixel with least intensity and make a new image that is black channel image, after we took the brightest pixel from the black channel image, considered as atmospheric light estimation .Each pixels of black channel image is divided by the atmospheric light and form a new image called transmission image and finally we multiply pixels of original image with the transmission map correspondingly, in result we get image without haze in grayscale. but we need the result in rgb form (colored image), we are not able to solve this problem, help me please
Solved! Go to Solution.
05-28-2023 10:56 AM
See if this can give you some ideas.... (Still no toolkit needed).
(I am sure the math is not exactly right, so modify as needed. I am sure things can be optimized more too)
05-28-2023 12:03 PM
05-28-2023 01:22 PM
As I said, the algorithm is not exactly as described, but you should be able to implement any other transform. The brightest pixel is of course from the headlight, and it is 255.
Many improvements are possible, of course. If you change the constant to 128, you get better brightness, but blooming in the headlight. You probably want some nonlinear mapping based on a lookup table (not shown).
05-29-2023 07:35 AM
Thank you , pixmap approach helped a lot.
05-29-2023 10:47 AM
Please help other users of this Forum by marking @altenbach's recent Posts "Accept as Solution".
Bob Schor