banding in my gradients

Effectively you will have banding in theory. But in practice the step differences in an 8 bit file (256 steps of brightness) are enough to make it nearly indistinguishable to the eye.

What the extra data DOES do is provide more data for algorithms to make guesses.

Think of it like this. On an 8bit file if you have a perfect black to white gradient and then adjust the levels so they that they fill the bottom 1% of possible brightness levels, then adjust the levels so they take up 100% again, you'll end up with a gradient bands of 2 colours, possibly 3 with rounding errors.

Do the same on a 16bit file and you'll have a gradient bands of 655 different colours. Now when you convert this to 8bit you'll still have 255 accurate colours because all that information that was lost in your conversion from 16bit to 8bit was not relevant.
 
Banding is a common problem in digitally produced gradients. The general consensus is that he transitions are too perfect, and RONDAL is indeed correct, adding a small amount of noise will alleveiate the problem.

I used to see this frequently with my commercial work when I first transitioned from film to digital, now I add a tiny amount of noise to the image and the banding magically disappears.

Mike Collete at Betterlight even makes a note of this effect, it may have been either in a tech note or a personal email, I don't remember at the moment. The Betterlight website is a good resource for those that strive for color accuracy and tonal fidelity. (lots of good info on art repro, commercial and fine art photography and color management)
 
I also remember hearing that our eyes are the most sensitive with gradients in the blue spectrum.

Actually you got that backwards. The eye is only sensitive to blue when no other colours overpower it. For proof of this, save a low quality (but still somewhat decent looking) JPEG, open it in photoshop and then compare the green channel to the blue channel :)
 

Most reactions

New Topics

Back
Top