[derwood]

Let's say you have a scan or photograph at a 16-bit instead of 8-bit pixel depth. How do I capitalize on the extra information when I convert it to 8-bit? The situation is that I have a 16-bit image that is physically too small for the final printed piece. But since it has that extra detail, I figure I should be able to almost double its dimensions without degrading the quality if I convert it to 8-bit. If I just convert the mode from 16- to 8-bit, it seems like all that extra data is just lost. Any insight? Thanks

[Philbert]

Bit depth doesn't work like that. The advantage of working with a 16 bit image is that you have much more "wiggle room" when tweaking the image. (Every time you alter brightness or color, you're throwing information away. Since 16 bit images have much more information to begin with, the penalty is much less severe.)
Think of bit depth as vertical information - up and down. What you need for your larger dimension requirements is more horizontal information (ie - more resolution). Completely different than bit depth. That said, if you're forced to upsample to meet the larger dimension, might as well do while in 16 bit and then convert to 8 bit. (although I doubt you'll gain anything)

[MacGizmo]

as Phillbert already said, bit depth has absolutely nothing to do with the resolution of the image when sizing it.
However, when you're sizing up an image, I recommend enlarging only about 10% at a time and running an Unsharp Mask filter between each increase in size. It helps keep the blurriness down.

[moose_n_squirrel]

In reply to:

---

However, when you're sizing up an image, I recommend enlarging only about 10% at a time and running an Unsharp Mask filter between each increase in size. It helps keep the blurriness down.

---

I was told, or read somewhere, that if you are using the Bicubic Smoother scaling in PhotoshopCS you no longer have to put yourself through all the 10% steps.

[MacGizmo]

That method does improve the situation, but it's not a total fix/cure.

[phlexrayon]

As long as you're on the upsampling idea, any thoughts about the Genuine Fractals software. Our studio uses it, but I'm not completely sold.
I have always used the 10%, unsharp technique and have found it to produce just as good quality as the software.

[derwood]

This has all been very helpful, thanks. I've never understood bit depth, but it makes sense now.
I'm going to confess more ignorance here though -- I understand the upsize and unsharp mask concept, but what/where is Bicubic Smoother scaling?

[charlieartist]

What about pxlSmartScale from Extensis? I have used it a bit, and found the results to be pretty decent.

[MacGizmo]

I've used both products and quite frankly, they work pretty decent (though I think Extensis works better) if you have a good size/resolution photo to begin with. So if you're wanting to take a 10x12 300dpi image and blow it up to 30x36, they work great. But if you have a 3x5 100dpi image and plan to blow it up to 11x17 or so, don't be surprised when the results are horrible. The old saying "cr@p in, cr@p out" comes into play.

[charlieartist]

Oh, how right you are.
Sometimes when that is the unavoidable course, though, I will deliberately run treatments, such as posterization, noise, Gaussian blur, to give it a more graphic treatment. Personally, I find this preferable to having an ugly photo with jaggies or JPEG artifacts...

[Philbert]

My experience is that the commercial scaling apps really don't do any better at up-rezing than the tools available in Photoshop. Especially with CS's Bicubic Smoother. Giz, I agree it doesn't completely replace the step methods, but it's quite an improvement over the previous resampling algorithms. (derwood - it's a new resampling option in PS CS that resides in the Image Size dialog.) And yes, how well ANY of them do is absolutely dependant upon the quality of the original image. (An image which needs little tweaking will fare much better than one which is squeezed and stretched to the limits - which is where a 16 bit image can help.)
Another thing that needs to be considered is that quite often, a larger image doesn't NEED as much resolution as a smaller image simply because of the greater viewing distance. (You view a poster from much farther away than an 8x10 print.) The typical billboard has a resolution of 8 - 12 PPI! If you were look at it from a foot away, you could actually count the number of pixels but from street-level distances, it looks fine. Also, it can depend on the output device. I've printed 30 x 40 posters from a lowly (by today's standards), 3 megapixel Canon D60 that looked great printed to a Displaymaker printer from a 72 ppi file. (No resampling involved!) But again, it depends on the output device. For inkjet, I'd go no lower than 150 ppi for large(ish) prints.