Fotoartist, you asked or made three good poiints. Here are my responses.

As physicists, we often try to look at limiting cases. So, I had the initially had borders on cells on a webpage in order to show that under those conditions it is tough to make distinctions between colors that are one-step apart in 5-biit color. But it also shows us that 5-bit color, while almost doing the trick, is not quite what we want. And this also shows us that 8-bit color is definitely more than we need.

As to doing my experiment in blue background colors in a table instead of green, here is the result. Note that banding does not occur. It is always better to do the experiment that just stating an opinion. :)

We do not need 16-bit color as I have shown by my many experiments here. If under only the most controlled and ideal conditions can a human see 10 million colors, i.e., that is 8-bit color in three RGB subchannels, then moving the minimum color step size from 1/256 to 1/(65,536_will not help us see more colors. All that moving from 8-bit to 16-bit gives us then is more colors that look the same. --Richard

Fotoartist, you asked or made three good poiints. Here are my responses.

As physicists, we often try to look at limiting cases. So, I had the initially had borders on cells on a webpage in order to show that under those conditions it is tough to make distinctions between colors that are one-step apart in 5-biit color. But it also shows us that 5-bit color, while almost doing the trick, is not quite what we want. And this also shows us that 8-bit color is definitely more than we need.

As to doing my experiment in blue background colors in a table instead of green, here is the result. Note that banding does not occur. It is always better to do the experiment that just stating an opinion. :)

We do not need 16-bit color as I have shown by my many experiments here. If under only the most controlled and ideal conditions can a human see 10 million colors, i.e., that is 8-bit color in three RGB subchannels, then moving the minimum color step size from 1/256 to 1/(65,536_will not help us see more colors. All that moving from 8-bit to 16-bit gives us then is more colors that look the same. --Richard

Thanks for the reply. Are you telling us not to believe our lying eyes when we see banding in a large smooth sky area in our photos? I am not a proponent of 16-bit color and almost never use it, but to me this is its main purpose in photography to prevent banding and posterization in large smooth areas.

Your color strip is too short, too pure in hue, and it has not been processed with a Levels or Curves adjustment that photographers love to use. Try that and viola!, the banding will be evident.

If 16-bit colors are necessary for one to do the editing they prefer, then they need to use it. All my demonstrations show is that any banding that comes through using editing software is not due to 8-bit color but due to the parameterizing of the software adjustments. I do not use Photoshop Curves or Levels and won't use them due to these errors they introduce.

If my original photo does not have banding in the sky or other places, then why would I use software which adds banding? --Richard

If 16-bit colors are necessary for one to do the editing they prefer, then they need to use it. All my demonstrations show is that any banding that comes through using editing software is not due to 8-bit color but due to the parameterizing of the software adjustments. I do not use Photoshop Curves or Levels and won't use them due to these errors they introduce.

If my original photo does not have banding in the sky or other places, then why would I use software which adds banding? --Richard

I propose an experiment.

Take a photo that includes a clear sky. Take the photo and save it as a tif or bmp to avoid any compression effects. Take the tif and convert it into an array of 16 bit integers.

Now shift the bits down 1, 2, 3..., 11 bits. That gives you 11 examples ranging from 5 bit to 15 bit integers. Multiply them all back up to 16 bits. That gives you the original brightness at the highlights but only 5 to 15 bit resolution. Convert the numeric array back to a tif or bmp. See what they look like.

Alternatively, just take the lowest 1...11 bits and make them zero.

I will do this experiment BUT as I am flat out trying to clean out the house for sale, I will not have time to do this for at least a week or two. If someone beats me to it, so be it. My thought would be to do this with Python, as I have done numeric analysis of several photos using that language and it seems to work well. I would expect that there are other ways to do this, that may be more appropriate to someone else.