Posterization - Is it always bit depth? (Long)
2002-05-28 by Todd Flashner
In the "8x16 bits and BW" thread I suggested it's possible that the posterization we sometimes see in our prints, contrary to conventional wisdom, may NOT be from "overworking a file", "breaking an image", "insufficient bit depth", "lost tones", or any of the various expressions used to suggest a problem that may occur from editing 8-bit files. Austin has recently sent me Piezo prints for my perusal, as I am interested in assessing that output, being the longtime Roark/MIS user I am. He has graciously allowed me to scan and post an example that SUBTLY illustrates the type of phenomenon I referred to. I've also encountered it in my workflow and it has stymied me because 16-bit evidence did not help. It's perfect this print comes from Austin because he does no 8-bit editing, and his histograms have no gaps within their endpoints. As such, any posterization in the image could not be due to the "lost tones" associated with 8-bit image manipulation. Furthermore, Austin's workflow excludes some of the other usual suspects, like interpolation, USM, or selections/masking. If there is such a thing as Certified Organic files that's what Austin prints from. So what we have is a nice print of a lovely image and what I will point to is not meant as a criticism of any kind - on the contrary, I think it may be so common that, to many of us, it goes unnoticed. I'm posting two samples of a scan of Austin's print, one a small jpeg of the overall photo, and the other, a TIFF detail of the area in question. The print was scanned in grayscale mode at 300dpi on my Umax Powerlook III. The files are in the Message Related folder for 5/02. Hopefully this link will get you there: <http://groups.yahoo.com/group/DigitalBlackandWhiteThePrint/files/Message%20 Related%20Files/05_2002%20uploads/> What I am addressing are the deep 3/4 tones along the right side of the woman's body. They extend from her armpit to her leg, and from the the bottom of her forearm to her hand. To my eyes the transition between tones in those areas is harsher and more abrupt than other tones in the print, and just have an unnatural appearance to me. I'd say the main characteristic is an area of mostly "flat" shadow tones which are "blocky" at their perimeter. Whatever it is, I suspect many of us experience it and refer to it as posterization. My sense is the effect looks more prominent in print than on screen. It's possible, but I have not measured it, that the ink color for that ink position is "cooler" than the other inks, which causes it to stand out against them, but since these scans originated in grayscale mode they would not reveal that. I'm interested to know: A) Do these tones look "natural" to you? Am I suggesting there is a problem where none really exists? (FWIW, Austin knew immediately what I was referring to when I mentioned it, so if you do feel I'm making something from nothing it's possible that it is more evident in the print than on the monitor.) B) If not "natural", do these tones look typical to you? IOW, do you get them with your workflow and materials? Conventional materials, or digital only? At what stage of the process do they appear in your work, ie, in the neg, scan, or print? C) Do you notice them using some materials or workflows more than others? My suspicion is that those areas in a well printed glossy silver print would have a much smoother feel/gradation, or a more natural placement of values. I'm guessing that those tonal transitions are somewhat present in the negative but are exacerbated by scanning, and then further amplified in printing. But that's just my hunch, I'd like to hear what others think. This print was produced with the Piezo driver using Piezo inks, by an experienced user. The Piezo product is considered my some to be the "gold standard" for smooth output and tonality, so how likely is it to be an effect of that process? I don't know what profile Austin used (for what I believe is Epson HeavyWeight Matte), but I do know his step wedges and Tyler's Zees prints looked smooth and uniform. One suspicion of mine is that this might be an effect off CCD scans. In the small sampling of drum scans I've had made for my negatives (thanks Mike K) I believe that the drum scanner handled such tones slightly better than my Leaf. Have any of you who've used both CCD and drum scanners noticed this occurrence more in scans from one scanner type over the other? The Leaf scans in 14-bits, same as most drums, so I don't think it's bit depth related, but, could the the sensor element, lens/aperture, or light source make a difference? Or, could this be a function of "grain aliasing" showing more prominently in those tones than others? I ask because their boundary seems blockier than other areas. On the other hand, it may just appear that way because the tonal transition is more abrupt, giving higher contrast at the perimeter. Could it be a function of too much separation of tones? IOW, is this information in the "toe" of the film's response curve and, as such, flat or compressed to begin with? In a silver print might get this area get further compressed into dark shadow and print much darker, and therefore go unnoticed? (I tend to think not since with my images I get the same effect in dark sky image areas of my MIS prints, but on silver papers these tones print with smooth, subtle, gradations). Obviously, I have more questions than answers. I believe this type of posterization (and/or whatever this is) is a somewhat common component of digital printing that was rather uncommon in conventional printing. Unfortunately, it's not always easy to determine where it comes from. Sometimes it's from overworking an 8-bit file, but I suspect, as often as not, it's from something else. I'm interested to hear what some of you think that "else" may be... Todd Flashner