Re: OT: 16 bit editing myth or reality?
Camden, I don't have the words to say how thankful I am about your reply. It made me realise how blind and deaf I can be sometime. Adobe photoshop marketing guys did such a good job leading us to believe they are number one, that every thing they say and do becomes the GOD SPOKEN TRUTH, we are basically brainwashed to believe it. What fool would dare challenge them, amoung other stuff, they created a tool to allow us to change the tonal relationship of our images and thank god they have given us not one but two ways to do it, one is simply ridiculous as it is, the hand tool with no simple way to edit the "curve" for precise adjustment, most likely nobody uses it anymore and a second one which gives us up to 16 control points that we can set and edit at will, the tool automatically ajust the in between points such that all is nice and smooth, cool ain't it. This tool is so great that it invariably creates hole in the histogram especially with 8 bit depth images, (that's because we can see them) with 16 bit depth images the number of levels is so large, they have to group them in order to show something that looks nice (and there isn't enough pixel on the width of the screen), the hole(s) (if any), are somewhere within these groups and we don't see them of course. This tool, is included in all of Adobe image editing application (I assume) and since Photoshop as basically taken up all the market that practically noboby would ever think there are other ways to do this, including some that don't create holes in histograms. I fell for it, I could even add that I fell head first in it and unfortunatly, probably most if not all of you did fell for it as well. The solution is so simple even I shouldn't have difficulty explaining it, you need to be able to create and edit all 256 levels and set both input and output values to integers. With a minimum of practice you can apply practically any so called "curve" you like and not create any holes, there is another side to this and I'm sure someone will pick it up. The software I use as that feature amoung others. Another very important point, having holes in an histogram is not the end of the world (unlike popular belief I suppose), if you are missing level 23 for example and still have levels 22 and 24 the contrast ratio becomes 256:(24-22), 256:2 or 128:1 and if I recall correctly 7 bit was plenty enough such that we wouldn't be able to distinguish such a small step, right?. We are speaking about multi megapixel images and these pixel levels could windup to be side by side, yes but also anywhere else on the whole image area where you haven't the most remote chance of being able to percieve it. Is it possible with Photoshop to create large enough holes that it becomes very visible, absolutely, that's basically what posterisation is but to create a good looking image, I don't think you'll go that far on purpose and one would find work around to those Photoshop limitations. In a curious way, the Photoshop solution to this problem is simple, just work in 16 bit mode, in the end you'll loose a bit less. It didn't take long for this to become THE GOOD WORD. I'll admit Photoshop can't do this mapping, lookup table or tonal adjustment curve in an easy way but it CAN even in 8 bit mode and it's with the hand tool, don't say it to loud, it can create a file you can save just as the other method and if you can find a way to edit these file, they have the magic number of control points, that's right, 256 points, you can even do this in up to 4 channels, ain't this magic. Another secret, I know of a tool you could use to edit such files but it works in Excel what a shame. It's also a shame that I didn't think of this before. Camden, I went to read the reference you gave me below and it says basically what you and everyone else is saying, I know all of this by heart, I even said it myself not so long ago. When working within the limitations Photoshop imposes on you it is true, all of it, maybe I should say most of it but anyway. It's mathematical, whenever you work with integers, if your not very carefull you are bound to loose a few and with Photoshop it's complicated to be very carefull and it becomes inivitable that you loose some levels (integers). I hope this settle the question of holes in histograms, if not then I'll be happy to give you more details. I would think the next question is about "good looking image", I know this is subjective but I assume an unwillingly posterised image wouldn't qualify as a good looking image wouldn't you agree? Is it possible to work in 8 bit mode and create excellent work? Today, you might be tempted to say no but not so long ago working in 16 bit mode was only a dream. How long do you think it will take before working with silver based negatives and papers become forgotten history? Does this mean the work of all the great artist is of poor quality? If you think it's hard to create a good looking image in 8 bit mode don't even think of doing silver base work, that truly difficult and very very slow. And now, "la piece de resistance" the heart of the problem if you prefer: If you allow your mind to believe that it is possible to work in 8 bit mode and create a good looking image. However easier, faster, more fun, etc. it could be to do the same thing in 16 bit mode, it would be extremely difficult to tell which is which ounce printed. Even in Photoshop as it is, converting the 16 bit image to 8 bit and then comparing numerically each pixels of an image to its counterpart, this is what the apply image substract actually does, the result would show minute differences but you would need at least 100% magnification to see them. Another thing you have to consider and maybe the most important, lets analyse this substraction thing in detail. When you do this in Photoshop, it (PS) take the value of the top most, left most pixel in image A and substracts it from the top most, left most pixel value of image B and so on, this means we have both the difference and location of these differences, substractions or errors. If the two image are identical the result or error would be 0 at any location, if the two image are not identical the error can range from -255 (0-255) to +255 (255-0) at any location. Now if the images are similar the difference at each location should be relatively small, maybe some errors will be -maximum error, ..., -3, -2, -1, 0, 1, 2, 3, all the way to the maximum positive error, this is why I suggested an offset of 128, to take care of the negative results, (PS) would set them to 0 and that's not what we want, it wouldn't be fair, right? These errors are also spread all over the resulting image area. We have already said that it would take an error of at least 3 to be become perceptible, an error of 0 is basically invisible at any magnification, an error of 1 (256:1) may become visible for a few people at quite high magnification, an error of 2 (128:1) will be visible to about half the people at at least 100% magnification and an error of 3 or above will be visible to practically everyone at 100% magnification. All this above is not what I say, it is what is generaly accepted by those who study these thing in detail or scientist if you wish. Now, if you are reasonably carefull while editing, the 8 bit edited image and the same but 16 bit edited image, the result should be fairly similar wouldn't you agree? Considering #1 we have 5 error levels that are practically invisible (126, 127, 128, 129,130, with an offset of 128), #2 the visible error levels are spreaded all over the test image area, remember this is a multi megapixel image, the visible errors on the test image become burried in those million of pixels and then for all practical purpose they become invisible under normal viewing conditions of our two image side by side. Those this mean 16 bit is a total waste, of course not, it's the best thing that happened in the image editing world, it is fast (if you have the computer for it), it is easier then working in 8 bit which would take a lot more effort to arrive at the same level of quality and you can do thing in 16 bit mode that you couldn't even do in 8 bit, 16 bit editing is the way to go, no question about that. Only a "fool" would think otherwise, why do you think I'm using PWP for my work, I can't afford PS for one and PWP as had 16 bit editing way before PS and that is for all of the tools not only a small portion of them as with the version of PS(7) I have. But but but, this doesn't change the FACT that one can do excellent work editing in 8 bit mode and that doing the exact same thing in 16 bit with the same image while easier, faster, etc. doesn't provide any visually perceptible benefit. Happy Holidays to all Yves ----- Original Message ----- From: "Camden Hardy" <camden@hardyphotography.net> To: <alt-photo-process-l@usask.ca> Sent: Wednesday, December 19, 2007 11:23 AM Subject: Re: OT: 16 bit editing myth or reality? > > On Dec 19, 2007, at 7:36 AM, Yves Gauvreau wrote: > > > What I basically say is this, if you take a normal multi megapixels > > image > > and edit it such that it looks good in either 8 bit or 16 bit mode, a > > "normal" human under "normal" viewing conditions wont be able to > > tell which > > is which. > > Not exactly. The point I was trying to make is that if you were to > make a straight print without any adjustments whatsoever, you're > right: no one would notice a difference. However, if any tonal > adjustments are made, everything changes. > > We've established that gaps in the histogram occur no matter what bit- > depth you're working in. The main difference is that 16-bit has the > capacity for exponentially more tonal values (256 times more than 8- > bit, to be exact). This allows these gaps to very few and far > between. 8-bit has much larger (and more) gaps after an adjustment. > I've seen this in my own images on several occasions; believe me, this > can happen during "normal" image preparation. > > The image itself shouldn't have any gaps in its histogram immediately > after capture. Only after making adjustments will these gaps > materialize. So if you convert to 8-bit right away you're exposing > yourself to the possibility of getting a poserized image. On the > other hand, if you wait until just before printing to convert back to > 8-bit (or even let the print driver do the conversion for you), the > gaps in your histogram will be much smaller, leading to much smoother > tonal transitions. > > I suggest you carefully read the article I posted; it does a much > better job explaining this than I can. Here's the link again: > http://www.earthboundlight.com/phototips/8bit-versus-16bit-difference.html > > > Camden Hardy > camden[at]hardyphotography[dot]net > http://www.hardyphotography.net
|