Some very basic questions about using an Epson V700 Photo please.

[Jiri Vasina]

For the end result picture, yes the 24bits are enough. But it's the modification of the images that is the crux. If you just scan the image and then print/present on web, it's OK to do it at 24bits. But if you need to modify the image it's better to do it in 48bits.

You may arrive at the same result if you do the work either in 24 or in 48bits, but more likely the result would be subtly different (the more changes, the more likely the difference).

And you are right, the visible color at (for example) 51.5% of the scale is the same in 48bit and in 24bits, but it's about the "graininess of the scale". If you have more smaller grains, the transitions are more smooth than if you have less of larger grains - is it not one of the reasons why we shoot LF? The smoothness of tonalities, of the transitions? Because the grain is relatively smaller in the comparison with the size of the image as a whole.

[Greg Lockrey]

For the end result picture, yes the 24bits are enough. But it's the modification of the images that is the crux. If you just scan the image and then print/present on web, it's OK to do it at 24bits. But if you need to modify the image it's better to do it in 48bits.

You may arrive at the same result if you do the work either in 24 or in 48bits, but more likely the result would be subtly different (the more changes, the more likely the difference).

And you are right, the visible color at (for example) 51.5% of the scale is the same in 48bit and in 24bits, but it's about the "graininess of the scale". If you have more smaller grains, the transitions are more smooth than if you have less of larger grains - is it not one of the reasons why we shoot LF? The smoothness of tonalities, of the transitions? Because the grain is relatively smaller in the comparison with the size of the image as a whole.

Let's not confuse the issue of "graininess" in terms of variance of color. It is more like "steps" of transition from one value to the next. Moving from step one to two in a 256 step wedge is indiscernible to most eyes let alone to ink-jet printers. I won't argue that the steps are finer in 48 bit but in practical matters the question is what is really necessary? Depending on the paper and ink most printers lose it from 1 to 20 and from 245 to 255 in terms of discernible value anyway and a keen eye can pick up about 2-3 step differences in between. In practical terms for example a 2400 spi scan of a 4x5 transparency in 48 bit takes about 20 minutes to accomplish verses about 5 minutes in 24 bit and for printing purposes not really have that much difference in the over all result. Plus the file size difference is significantly different to load into the printer as well. Time being money, is the perceived improvement worth the extra cost in time? I bet that only a handful of people can tell the difference.

[Jiri Vasina]

I used the word "graininess" solely on the steps in the color wedge. It was not meant to relate to graininess of the image (that was only an example). In hindsight the word was not chosen well, but it sounds better than "discreteness".

I completely agree that human eye can not discern 256 shades of grey (or any color). But have you never encountered those irregular blue banding in pictures of skies? The banding is not there if the image is 48bit, but suddenly appears in 24bit. And not only on monitor, but also on the print? The reason being that even the 256 steps are too discrete, not smooth enough.

Anyway there are horses for courses - sometimes you get the same result with two very different beasts (48bit vs. 24bit). And sometimes having a race champion (48bit) will only slow you down.

[Greg Lockrey]

I do know where you are coming from, Jiri, I use 48 bit 90% of the time myself. I was trying to keep the terminology clear for anyone else reading this thread. But there are many applications that can't use it (like stitching in older versions of Photoshop for example) so I keep those operations in mind whenever I do my scanning. I routinly make scans of 30x40" artwork on a 12x17" flatbed and I do a lot of stitching. Keeping files just large enough to be good enough is a priority when it can take over an hour to make those kind of stitches and I have a pretty powerful computer which is another important consideration in the mix.

[Wally]

Let's not confuse the issue of "graininess" in terms of variance of color. It is more like "steps" of transition from one value to the next. Moving from step one to two in a 256 step wedge is indiscernible to most eyes let alone to ink-jet printers. I won't argue that the steps are finer in 48 bit but in practical matters the question is what is really necessary? Depending on the paper and ink most printers lose it from 1 to 20 and from 245 to 255 in terms of discernible value anyway and a keen eye can pick up about 2-3 step differences in between. In practical terms for example a 2400 spi scan of a 4x5 transparency in 48 bit takes about 20 minutes to accomplish verses about 5 minutes in 24 bit and for printing purposes not really have that much difference in the over all result. Plus the file size difference is significantly different to load into the printer as well. Time being money, is the perceived improvement worth the extra cost in time? I bet that only a handful of people can tell the difference.

24 bit is 8 bits per color, three colors, right? And 8 bits is only 256 discrete hues of each of these colors.

If your image has an area with a gradual change in hue and the hue is close to one of the three colors, in 24-bit mode you're probably going to see some banding. With 16 bits per color in 48-bit mode, there are 64K hues for each of the three colors and therefore 256 gradations across the transition of that band, and it's much less likely that you will see any banding.