Ad
Ad
Ad
Pages: « 1 [2] 3 4 »   Bottom of Page
Print
Author Topic: 16bit printing...your thoughts?  (Read 10664 times)
hjulenissen
Sr. Member
****
Offline Offline

Posts: 1677


« Reply #20 on: May 04, 2011, 01:08:55 AM »
ReplyReply

Hi Alain,

Presumably, yes. One can only hope for the sensible thing being done, first a colorspace conversion then followed by a rounding down to 8-b/ch data in case of an 8-b/ch pipeline. However, when starting with an 8-bit/channel source that is an opportunity lost. The only thing left to somewhat mask the potential posterization is clever dithering or adding noise.

Cheers,
Bart
Slightly philosophical here:
The final image could be thought of as 1-bit (when doing B&W): either a piece of paper is covered in ink, or it is not. Perhaps in reality it is >1 due to combination of ink etc, but my point still stands that the final print covers a minimal range between maximum black and maximum white.

What you call "clever dithering and adding noise" is therefore needed, as the spatial resolution is very high, while the tonal resolution is low. By encoding any given 8-bit pixel as a number of ink/no ink "pixels" on paper, the viewer will get the impression of smooth tonality.

Since spatial/tonal resolution is decoupled this way, it makes me wonder if matching some exact dpi-setting or bit-depth for printing from photoshop should really be needed in an ideal world. In the real world, driver and hardware manufacturers make choices that are less than ideal, of course :-)

-h
Logged
madmanchan
Sr. Member
****
Offline Offline

Posts: 2108


« Reply #21 on: May 04, 2011, 09:58:24 AM »
ReplyReply

Generally I recommend keeping the image in 16-bit through the entire workflow that is in your hands. For example, if printing from PS, print from the 16-bit document. That means PS will perform the color xform from the document space to the output space (via the ICC printer color profile) with >= 16-bit precision. If the final output stage requires 8 bits, then PS can drop the processed image down to 8 bits (with dither). LR also does this.
Logged

Doombrain
Full Member
***
Offline Offline

Posts: 206


WWW
« Reply #22 on: May 05, 2011, 06:41:04 AM »
ReplyReply

Epson large format only support 16bit in Mac OSX, however as DD says there's very little benifit in 16bit printing at the moment
Logged
Scott Martin
Sr. Member
****
Offline Offline

Posts: 1311


WWW
« Reply #23 on: May 05, 2011, 07:20:54 AM »
ReplyReply

Epson large format only support 16bit in Mac OSX, however as DD says there's very little benifit in 16bit printing at the moment

Doombrain, see my post on the previous page. Canon is the only one admits to having high bit depth data processing on their printers, and their printers do, in fact, show some benefit. While many drivers are now capable of sending 16 data to the printer, there's some debate about what other printers are capable of processing that data at a high bit depths. RGB to 8-12 color conversions and dithering takes place on the printer itself (not the driver on your computer). To do all of this on the fly at 8 vs higher bit depths requires serious computational power and cost.
Logged

nemophoto
Sr. Member
****
Offline Offline

Posts: 486



WWW
« Reply #24 on: May 05, 2011, 10:17:26 AM »
ReplyReply

Ditto on the comments about the Canon iPF printers. They do implement 16-bit printing well. However, in my experience, it's mostly seen when you have a very long, subtle tonal graduation (large expanse of sky) and especially for B&W. Otherwise, I find there are times the 8-bit driver is just as effective. I know this doesn't help you much with your Epson dilemma, but I don't think Epson has truly implemented 16-bit printing the way Canon has. Therefore, I don't think you'll see a lot of difference.

As for the Mac vs. PC debate, don't let yourself be complacent. Mac Malware is out there and expanding. Up to now, PCs were the target because there are so many of them in use -- bigger bang for the buck. But with more Macs being sold, it's a larger target audience (ditto for iOS). Here's an interesting blog from ZDNET: New MAC OS X scareware.

Personally, I use PCs. A lot more bang for the buck. My wife, with 25-yrs as a graphic designer, has used Macs for years and is having a tough time ponying up $3500 for a new computer that I could buy/build as a PC for her for about 1/2 the cost. And honestly, once you work in InDesign, Photoshop, etc., there's basically no difference. And let's face it, how much time do you REALLY spend with an operating system (yes, it's the underpinnings). However, since we share an office, I know my wife has at least 2 or 3 program crashes or system lockups A DAY. A lot more than I have with Win 7  x64.

My 2-cents.

Nemo
Logged

Sven W
Sr. Member
****
Offline Offline

Posts: 514


« Reply #25 on: May 05, 2011, 01:41:26 PM »
ReplyReply

Canon is the only one admits to having high bit depth data processing on their printers, and their printers do, in fact, show some benefit.

I think it was back 2002 ImagePrint introduced a true 16 bit pipeline.
I still use IP on four Epson LFP and always tries to work and print everything in 16.

/Sven
Logged

Stockholm, Sweden
Doombrain
Full Member
***
Offline Offline

Posts: 206


WWW
« Reply #26 on: May 09, 2011, 07:29:14 AM »
ReplyReply

Doombrain, see my post on the previous page. Canon is the only one admits to having high bit depth data processing on their printers, and their printers do, in fact, show some benefit. While many drivers are now capable of sending 16 data to the printer, there's some debate about what other printers are capable of processing that data at a high bit depths. RGB to 8-12 color conversions and dithering takes place on the printer itself (not the driver on your computer). To do all of this on the fly at 8 vs higher bit depths requires serious computational power and cost.

Thanks.

Epson printers have no processing on the printer itself apart from some basic RGB rendering. all rastering is done via s/w on the PC/Mac. As it stands >OS10.5 supports 16bit with Epson drivers and a few RIPs on Windows.

while there's improvement on gradations for proof work, there's little benefit in photo/art work IMO
Logged
Scott Martin
Sr. Member
****
Offline Offline

Posts: 1311


WWW
« Reply #27 on: May 09, 2011, 08:39:30 AM »
ReplyReply

Epson printers have no processing on the printer itself apart from some basic RGB rendering. all rastering is done via s/w on the PC/Mac.

All printers do a fair amount of on-board processing. The printer drivers send RGB data and the on-board processing does the RGB to 8-12 "N" color conversion along with the screening/dithering rendering (the final rastering within these N color channels). The conversions and dithering are major intellectual property that they want to hold close to their chests so to speak. In some rare cases, OEMs have signed agreements with RIP manufactures to share that processing which can speed up the process. Without that onboard processing RIPs have a lot more work to do and send far more data to the printer.
Logged

digitaldog
Sr. Member
****
Offline Offline

Posts: 9011



WWW
« Reply #28 on: May 09, 2011, 01:34:38 PM »
ReplyReply

I tested the 16-bit vs. 8-bit differences for my iPF6300 in two ways:

1. Printed a custom 1728 patch target built in i1Profiler in 8bit and 16bit (driver set in both cases for highest quality). Measured and compared the deltaE differences using ColorThink Pro (the results surprised me):

dE Report

Number of Samples: 1728

Delta-E Formula dE2000

Overall - (1728 colors)
--------------------------------------------------
Average dE:   0.28
    Max dE:   3.30
    Min dE:   0.01
 StdDev dE:   0.36

Best 90% - (1554 colors)
--------------------------------------------------
Average dE:   0.18
    Max dE:   0.58
    Min dE:   0.01
 StdDev dE:   0.12

Worst 10% - (174 colors)
--------------------------------------------------
Average dE:   1.14
    Max dE:   3.30
    Min dE:   0.59
 StdDev dE:   0.55

--------------------------------------------------

All the high delta’s were in greens*! I think this illustrates that you may need to send the appropriate bit depth out target and image wise. Anyone else see anything like this?
*162.3   255.0   46.4   

2. Printed a very difficult synthetic image in both 8-bit and 16-bit: Bill Atkinson’s 28 balls test image. The color space was Adobe RGB (1998) and again, the driver was set for highest quality for both documents. Note also that Dither was turned OFF in Color Settings.

Results: I can see a difference, the 16-bit output has smoother graduations in many areas of the various balls. You need to look closely but its quite visible. This is a greater difference than I’ve seen in the past using the Epson driver set both ways. If you have a high bit file, you’d certainly want to pass it to the Canon export module and set it for highest quality (16-bit).
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Scott Martin
Sr. Member
****
Offline Offline

Posts: 1311


WWW
« Reply #29 on: May 09, 2011, 01:47:24 PM »
ReplyReply

Results: I can see a difference, the 16-bit output has smoother graduations in many areas of the various balls. You need to look closely but its quite visible. This is a greater difference than I’ve seen in the past using the Epson driver set both ways. If you have a high bit file, you’d certainly want to pass it to the Canon export module and set it for highest quality (16-bit).

Yes, that's exactly what I've been seeing from my tests since these printers first came out (2006?). I've been profiling them using true 16 bit targets since the beginning for that very reason. Good to hear your feedback!
Logged

digitaldog
Sr. Member
****
Offline Offline

Posts: 9011



WWW
« Reply #30 on: May 09, 2011, 01:58:18 PM »
ReplyReply

I've been profiling them using true 16 bit targets since the beginning for that very reason.

And yet the targets out of i1P are not 16-bit although I did convert em.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Scott Martin
Sr. Member
****
Offline Offline

Posts: 1311


WWW
« Reply #31 on: May 09, 2011, 02:00:24 PM »
ReplyReply

And yet the targets out of i1P are not 16-bit although I did convert em.

Seems like converting them from 8 to 16 wouldn't help much. I've actually painstakingly made my own 16 bit targets in Photoshop using true 16bit values for each and every color patch. I've also made my own reference files that contain true 16 it data.
Logged

digitaldog
Sr. Member
****
Offline Offline

Posts: 9011



WWW
« Reply #32 on: May 09, 2011, 02:04:45 PM »
ReplyReply

Seems like converting them from 8 to 16 wouldn't help much. I've actually painstakingly made my own 16 bit targets in Photoshop using true 16bit values for each and every color patch. I've also made my own reference files that contain true 16 it data.

There’s either a difference in how the driver handles the data or the data itself. I suppose I could try to subtract them in Photoshop and see (again, Dither is off). Photoshop doesn’t make this easy when the two doc’s are different bit depths.

So do we need to get X-Rite to export 16-bit data? I can’t imagine why its not the default anyway. Then the next question would be, would they do that and how darn long would it take them?
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Scott Martin
Sr. Member
****
Offline Offline

Posts: 1311


WWW
« Reply #33 on: May 09, 2011, 02:09:16 PM »
ReplyReply

So do we need to get X-Rite to export 16-bit data?

Yes, I think so! I vote for full 16 bit support for both for target generation and for measurement data and processing.

Then the next question would be, would they do that and how darn long would it take them?

Oh man, I'm not going to touch that one, LOL!
Logged

digitaldog
Sr. Member
****
Offline Offline

Posts: 9011



WWW
« Reply #34 on: May 09, 2011, 02:13:45 PM »
ReplyReply

OK, I asked in the proper channel for this so it seems reasonable we should see it any day now... Oh sorry, major brain fart.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Schewe
Sr. Member
****
Offline Offline

Posts: 5469


WWW
« Reply #35 on: May 09, 2011, 06:25:59 PM »
ReplyReply

I've actually painstakingly made my own 16 bit targets in Photoshop using true 16bit values for each and every color patch. I've also made my own reference files that contain true 16 it data.

Well, if you were using Photoshop then the target file ain't a full 16-bit. They are 15+1 bit targets since Photoshop doesn't do full 16-bit files...PS will only deal with 0-32,768 for a total of 32,769 levels in Photoshop...not 16-bit.
Logged
digitaldog
Sr. Member
****
Offline Offline

Posts: 9011



WWW
« Reply #36 on: May 09, 2011, 06:49:47 PM »
ReplyReply

Well, if you were using Photoshop then the target file ain't a full 16-bit. They are 15+1 bit targets since Photoshop doesn't do full 16-bit files...PS will only deal with 0-32,768 for a total of 32,769 levels in Photoshop...not 16-bit.

Further, I tried sampling a target up and down a number of times in Photoshop and comparing to the original. Using Apply Image>Subtract, the results appear to show no difference which kind of makes sense. I have dither off. The patches are solid colors. I’m not certain having the package save out the colors values in high bit would bring anything to the party but I’m open for suggestions in further testing.

Now there clearly is a difference in terms of what the target produces from a Spectrophotometer when 8-bit vs. 16-bit data is sent to the driver. But it doesn’t appear to be based on the target itself, at least as far as I can see with the limited testing I’ve done.
Logged

Andrew Rodney
Author “Color Management for Photographers”
http://digitaldog.net/
Schewe
Sr. Member
****
Offline Offline

Posts: 5469


WWW
« Reply #37 on: May 09, 2011, 10:04:45 PM »
ReplyReply

Further, I tried sampling a target up and down a number of times in Photoshop and comparing to the original. Using Apply Image>Subtract, the results appear to show no difference which kind of makes sense. I have dither off. The patches are solid colors. I’m not certain having the package save out the colors values in high bit would bring anything to the party but I’m open for suggestions in further testing.

As long as you have dither turned off, I wouldn't expect to seem much difference. However, you could clearly spec different color targets if you were starting from within PS using high bit-depth color specifications.
Logged
JeffKohn
Sr. Member
****
Offline Offline

Posts: 1671



WWW
« Reply #38 on: May 09, 2011, 10:14:00 PM »
ReplyReply

Quote
Now there clearly is a difference in terms of what the target produces from a Spectrophotometer when 8-bit vs. 16-bit data is sent to the driver. But it doesn’t appear to be based on the target itself, at least as far as I can see with the limited testing I’ve done.
I don't think 16-bit test targets would be of any benefit. First, unless your test targets have more patches than can be represented in 8-bit color, what is the point? Second, even if you did create such a target, I'm not sure our spectro's are accurate enough to register such minute differences reliably. Trying to measure so many closely-spaced data points could actually hurt profile smoothness. I've seen this when creating monochrome profiles using QuadToneRIP's  QTR-Create-ICC. I got smoother results by averaging multiple 21-step charts than I did by 1 (or mulitple averaged) 51-step charts.

Logged

Scott Martin
Sr. Member
****
Offline Offline

Posts: 1311


WWW
« Reply #39 on: May 09, 2011, 10:25:07 PM »
ReplyReply

Quote
They are 15+1 bit targets since Photoshop doesn't do full 16-bit files...

I'm going to miss that little piggy!!! Yes, I understand.

Further, I tried sampling a target up and down a number of times in Photoshop and comparing to the original. Using Apply Image>Subtract, the results appear to show no difference which kind of makes sense. I have dither off. The patches are solid colors. I’m not certain having the package save out the colors values in high bit would bring anything to the party but I’m open for suggestions in further testing.

Another test is needed. Try this: Take a BA 16 bit target. It's a true 16 bit file. Which means that a middle grey could 127.5 not just 127 or 128 - there are lots of increments in between 8 bit integers! :-] Take this 16 bit BA file and convert it to 8 bits and back to 16. Do your subtraction comparison (there are at least two ways of doing this - I personally like to have two layers with the top blending mode set to "subtract") and look at the numbers.

Do the numbers all show "0"? Well, think again! Click on the eyedropper in the info palette and switch it to show you 16 bit values (which are really 15 bit 0-32,768 values as Schewe brought up). There *are* differences - just differences so small that a crude 8 bit scale doesn't illustrate. Thus is the delicate nature of 16 bit imaging!

So, being a geek, I like to make custom made 16/15 bit (let's just call them "high bit depth") profiling targets and use them with corresponding high bit depth references files. Just like Bill (gotta give credit where credit is due). Tear apart my ColorPort XML files and you'll see numbers like Red=255.501 Green=127.533 Blue=33.434 instead of Red=255 Green=128 Blue=33. ColorPort's reference files also contain measurements to the third decimal.

True high bit depth profiling hand-in-hand with true high bit depth output seems a wonderful thing. I enjoy it, regardless of how small the benefit may be. :-]

Logged

Pages: « 1 [2] 3 4 »   Top of Page
Print
Jump to:  

Ad
Ad
Ad