Re: A 16 Bit vs. 8 Bit Comparison, authored by Todd Flashner, 6 January 2002.
2002-03-23 by Todd Flashner
Hi Austin, I hope you don't mind if I reply to your off-list post on-list, but I think you raise good questions that others have touched on and may like to weigh in on too. > A quick note/question on your write-up. It's a very good write-up, BTW. I > don't believe the clips you are using at the bottom contain the top part of > the histogram. You have a white shirt collar in the image, and that would > be quite a bit lighter than the eyes...so I believe that's the area that the > very top part of the histogram would be representing, right? Yes, the histograms were taken from the entire image. My thought at the time was to show what the entire images' histograms were, primarily to show the extent of the moves, then to select a full tonal range area to demonstrate visual parity. BTW, I did inspect the entirety of both images and the same parity held throughout. After your, Bill's, and Barry's posts, I see the histos of the details would have made more sense. Perhaps this will get me off my butt to expand the article and include that. Antonis, your grayscale idea was good too. It was the sloth in me that said "no". > Also, don't the JPEG images have all the tones, since JPEG will basically > re-size the image and fill in the missing tones? I think those images won't > show the posterization for both those reasons... Well that I suppose is another topic open to discussion and experimentation. All I can say is that the jpegs (high quality) fairly represent what the "images" themselves looked like with my eyes; that is to say, both were full ranged and with detail, with barely, if any, discernable difference between them. (If I might digress... what I find interesting is that if there is a discernable difference, I'll bet at least 50% will prefer the 8-bit version. It might look a bit sharper.) True, the histograms get rewritten, much as a resizing, rotation, or blur would do. To my mind that just furthers the point that for much image editing the histogram is NOT a good indicator of image quality, precisely because it is so easy to make a good histogram from a bad image. > What do you think? I histogramed those two images, and they both have full > histograms, up until they cutoff, not having the white shirt collar in > them...so even if the posterization exists in the .tiff files, it doesn't in > the .jpegs... The question then becomes, does jpeging actually make a bad "image" look better? IOW, if the IMAGE itself actually VISIBLY shows posterization will jpeging that image ameliorate the posterization, as it's smoothed histogram would suggest? It's a good question, worthy of some testing, and I don't honestly know the answer. My suspicion is that that you'd actually end up with the ironic situation of an image looking worse while the histogram looks better - but that's purely speculative on my part. > If you want a place to post the two .tiff images, you can use my FTP or web > site. Thank you. I think what would make the most sense would be to put the link to the site in the groups files section along with the two detail TIFFS, if Martin and Antonis feel we can afford the space. Otherwise I may take you up on that. Thanks, Todd