PhotoShop is feeble at 16 bits.

PhotoShop is miraculous in terms of what it can do. However, when one is trying to work with high quality images in 16bit mode. Only a hand full (a useful handful) of features is available. To do more sophisticated PhotoShoping you have to truncate to 8bits. This is probably fine for color but for black and white falls short of what the human eye can discern: at least 12bits of gradation (I would gladly be corrected on that piece of information). Does anyone have best practices for working with high quality Black & White scans and does anyone know of plug ins that allow the use of more filters, for instance, in 16bit mode?

Do everything you can in 16-bit color, and keep in mind that is 16 bits per channel, or 48 bit color total. Then convert to 8 bit/channel (24 bit) color to do whatever you need to do in layers. Then convert to B&W by using channels/monotone, not the grayscale (256) mode conversion. This way you preserve the color information and can adjust the overall image tonality to a very fine degree.

 

Cheers,

Britt,

 

Check out Ian's tutorial on PS 16-bit image editing at http://www.computer-darkroom.com/16bit/16bit_1.htm .

 

Joffre

Thans for the pointer to Ian's article. Very clever. I might have stumbled on it in a couple of years ;)

I disagree with you about whether or not 8 bits is adequate for b/w. It allows 256 different shades of gray, which is well beyond what the human eye can distinguish. Your problem may be more with how you are printing. Good b/w prints from digital sources require the use of special inks. I haven't made any myself, but it is my understanding that such prints can be superior to conventional darkroom prints.

 

Since I use the Gimp under Linux, I do everything in 8 bit except the basic scan. My b/w prints from scans are not as high quality as darkroom prints, but I use just the black ink in my Epson 1280. Still they are not bad. The main problem is that I can almost see the dots, something that doesn't happen with color and wouldn't happen with a better choice of inks. At first I had some problems with delicate differences in shades of gray, but with practice I've been improving. That, I think is just a matter of technique, and you have to work at it. When I get around to it, I'll probably do something about improving the ink set situation and then I think my b/w prints will be much better and as good as I would want.

The samew site has a good tutorial on the channels method of converting to B&W I suggested in my earlier post: <a href="http://www.computer-darkroom.com/tutorials/tutorial_2_1.htm">http://www.computer-darkroom.com/tutorials/tutorial_2_1.htm</a>

Leonard, if all you do is scan and print without any corrections then 8 bits is fine. As soon as you try to do any major adjustments then you start to lose quality. The thread on <a href="http://www.photo.net/bboard/q-and-a-fetch-msg?msg_id=004kNm">why image corrections work smoother after changing a 8bit scan to 16bit</a> goes some way to explaining this. Non-linear functions like gamma, curves, etc cause huge gaps in the histogram if not applied at scanning stage.

One more trick using the Duplicate Image techniques in Ian's tutorial that helps get around the fact that you can't use adjustment layers in 16-bit mode: you can write curves in an 8-bit copy of the image, save the curves, then apply them to the 16-bit image once you are satisfied with them. Since you can even do this in an adjustment layer on the 8-bit image, you can perform a good deal of experimentation without worrying about accidentally making a destructive change to your image. Once you apply the curves and save the 16-bit image, of course, the change is permanent (unless your 16-bit version is also a copy), but it's a handy trick.