Editting in 8 bits vs 16 bits !!!

HI fdear friends,
I've been reading a "scott Keby" book to learn how to better use PS CS, this guy is good at tips, but he reserved just less than one page explaning the advantage of editting in 16bits. I understood that it helps better manipulating colors and the advantage of RAW...
As usulal, it's in this forum that you get the best explanations or links,
so what do you guys have on this topic ?
thanks in advance
regards
Khalid

First of all, when we talk about 8 bits, we mean 8 bits per color channel for a total of 24 bits, and when we talk about 16 bits, we mean 16 bits per color channel for a total of 48 bits.

Next, in nearly all cases, the 16-bit file will be twice the file size as the 8-bit file. If you have oodles and gobs of hard disk, then that isn't so bad. Some people will use 16-bit through the entire editing process, then convert that file to 8-bit for practical use, then archive the 16-bit file away for safe keeping.

Some graphics programs will allow full and complete functions on 16-bit files. Some programs only allow partial functions on 16-bit, and everything else is 8 bit. Some programs will not deal with 16 bits at all, so the whole thing is 8 bits.

Obviously the 16 bit files will allow for a more gradual tone range in the image. If you were shooting a sunset, it would really look bad if you used only 4-bit color, and you know how much better it looks in 16-bit color? Well, imagine if you go to 16-bit color. It looks more gradual yet. On the other hand, some people don't see the difference. Most photographers will easily see the difference. Who are you trying to impress with your great photography?

---Bob Gross---

I've tested this once and posted results in some other topic and I concluded that for normal use, you don't really need 16 bits. Thing is, when you convert 16 bit photo to an 8 bit one, not two pixels had the same tone, like the top left corner pixel in 16 bit mode and the same one in converted-to 8 bit mode. It is obvious that there are less colors, but even on an A3 size printout it would be impossible to notice, plus, I am not sure the lab's machine even supports 16 bits for prints. However when editin a photo and adjusting saturation/brightness/​contrast it definetely is a good idea to do it in 16 bit mode if you're somewhat of a perfectionist. It is very hard to say how much it would affect an image if one was going 8 bit all the way and the other 16 bit, I would bet that very little, so little it can be practicaly forgotten. But soon computers will be a bit stronger, and HDDs even cheaper and PS will have full support for 16 bit graphics. That is when I'll kiss the 8bit mode good bye. Now I obviously use 16 bits in RAW converter, convert to TIFF 16 bit (only those which I intend to print), edit in PS until I can in 16 bit, convert to 8 bit, apply filters/whatever, save as 8bit tiff.

I tend to always convert in 16 bit (I dont think it truly is 16 bit, more like 12 in truth) even if I then ultimately output 8 bit tiffs. the reason is that you get a lot less 'gappy' histograms if you do the colour adjustments in 16 bit than 8. sure you may not always see the end results but extreme colour adjustments can start to look blocky in 8 bit.

I started using 16 bit before I had my DSLR. I used to scan my negatives and found that the trickier lighting conditions came up a lot better when working with 16 bit scans. I take a lot of pictures of aircraft and the exposure against the bright sky can upset your exposures. Some scans looked like silhouettes in 8 bit but in 16 bit the detail on the dark aircraft was there to see.

Since then, I have figured that it is easier to ditch data later than to try and recover what isn't there.

Note that virtually all output devices (monitors and printers) have no more than 8-bit precision. Having greater precision than this in your final product isn't really useful, I don't think humans can really perceive it any way.

Where >8 bit precision is useful (the Canon sensor only records 12 bits per channel) is in editing. If you're making significant changes to the image in post-processing, for instance making severe "levels" corrections to try to salvage a poorly exposed image, then preserving all of the detail of the original throughout the editing process is extremely important. In effect, you have a greater "diversity" of tones to work with, so that when you remap pixel values, you'll continue to have diversity of tones, with fine gradations, in the final output.

Of course, all this assumes that you've been shooting in RAW, and thus have 12-bit per channel precision to begin with. If you shoot JPG, then the camera has reduced the original 12 bits down to 8 bits, and subsequent editing or saving that image in 16 bits would be fairly pointless.
-harry

thanks guys for the precious explanations.
I recorded then that it's more about the gradual change in a defined color than about having more colors. I also understood that you have defenitely to shoot in raw;otherwise switshing from 8 to 16 bits from a JPEG file (having recorded 8 bits) is seemly useless.
so for printing out, do prints in general show the difference ? and how far can the printer quality matter in this subject.
I'll do more searsh about this and get back to you
regards
Khalid

In addition, I think you should be primarily concerned about differend color profiles and correct color calibration of your own monitor. I've had, and still have problems with printed color precision (so it seemed), on prints the colors are warmer and the images are overall darker. I've asked for their print machine PS profile (.ics if I recall) and got it, now I still have to do some monitor calibration I hope to borrow a colorimeter such as a Spyder.

I would beg to differ. Although I would advise someone to shoot in better-than-8-bit-JPEG mode in the first place, if you start with an 8-bit file, and edit it in an 8-bit space, you essentially have no overhead for image manipulations.

Consider that most image operations are multiplicative, that is, adjusting a level by one stop actually multiplies (or divides) the luminance values by 2, thus shifting the values one bit position to the right or left. If you have a fairly wide-range image to start with, say luminance values from 5 to 250, and do some masking operation where you are 'adding' two images together, you can quite easily blow out the hightlights or block the shadows, even before you do a final brightness or contrast adjustment on the result.

As far as the original book mentioned goes, I would be pretty suspicious of someone who purported to tell you all about how to edit images in CS, only to relagate all mention of 16-bit to a small chapter at the back of the book. Smells to me like a lazy book author (or a greedy publisher) that took a PS 7 book, and tweaked it a little for the CS release, and rushed it to market. I'd pass on that book, and look for something a bit more professional.

Part of the issue with 'how many bits' is the inherent compression of the image at every stage.
A brightly lit scene with deep shadows can have a brightness range of over 10,000:1. This corresponds to about 14 bits (and yes, I am greatly simplifying this).

A 10D , and most good digital cameras, allows a capture range of 12 bits, already compressing the image by 2 bits, or about 4:1 (probably more in reality)

A print can have a range of maybe up to 200:1, a little less than 8 bits.

Since we are progressively compressing the scene (and losing information) each step of the way, it makes sense to preserve as much range as we can in each step, using RAW to get 12 capture bits, and 16 bit software to insure no bits get discarded in processing.


I can see a difference in RAW & JPG in some images, no discernible difference in others. I play it safe and switch to RAW when I want the best quality.


boB

Ok, nice so far,
A question a lot of people ask : when you finish processing your raw picture and want to send it for a print, what is the best format to save it on ? mor can companies print stainght from raw files
/ cause I tried once with tif format and I realized that they switched it to jpeg before printing it out!!!!
thanks inn advance

I think it's a common misconception to try to associate bit depth with dynamic range in this manner. Given the linear mapping of luminosity to channel values in RAW files, you _could_ probably make some argument about the maximum possible dynamic range based on bit depth, but the real dynamic range would be much less due to noise (the "small channel values" are insignificant compared to noise, and so do not capture detail).

But when converting that linear 12-bit capture to jpg, the channel values are remapped in a non-linear manner, so the range of 0-255 could represent a dynamic range of a gazillion stops if you want it to, or just 2 stops, depending on how the mapping is done. In this case, the bit depth really just tells you how many distinct "steps" you have in between the extremes, or how fine your gradation of tone is across the range.
-harry

One of the risks of oversimplification (for the sake of keeping the message down to a reasonable size) is not providing enough detail for those who have a good understanding of the issue.

Expansion & compression of an image (to fit 8 bits) is ok as long as we are ok with the compromises made. Expanding a 2 stop image probably just wastes data but compressing 14 stops to fit 8 bits definitely loses data.

Since most (all?) printers, including chemical processes, work with 8 bits or less we must finally compress the image to fit. Much of the fun of photography is that we, the photographers get to choose where the compression takes place in order to show the scene in the way in which we intended.


boB

I've had that too, so I save as TIF or PSD, & convert to a max quality JPG if I'm going for $.20 prints at the drugstore. TIFs are best for most home printers I think.