Luminous Landscape Forum
Equipment & Techniques => Medium Format / Film / Digital Backs – and Large Sensor Photography => Topic started by: makercob on January 02, 2007, 01:12:15 am
-
I had been implementing 16 Bit TIFF in my workflow whenever possible.
But recently I did some tests, comparing 16 bit version and 8 bit version from one same P45 raw file. (from CaptureOne DB, Adobe 1998, ISO 50, IIQ Large, no exposure/contrast/saturation adjustment and no sharpening in C1)
In photioshop , both 16b and 8b images were adjusted by a same very steep curves.
Although their hisograms are definitely different, the difference in the actual images is so subtle, that it only become barely visible when zoomed beyond 200%.
Is the smoother gradient promised in 16 Bit TIFF overcomed by the CCD noise?
Is it is really worth using 16 Bit TIFF? 16b TIFF heavily slows down photoshop, (especially in PC systems, which have 2GB RAM limit for photoshop)
-
A few comments:
Your upper window is identified as RGB/16 in Photoshop. Mix up?
Secondly, the advantages of 16 bit mainly affect gradations. Here is a dark blend which has been processed using a steep curve:
(http://forums.rennlist.com/upload/8v16bit.jpg)
8-bit on top, 16 bit below. You can see that the 8-bit file produced a rougher transition.
-
Please ignore the upper window's title...
16 bit can produce a smoother transition in the test like you did, draw a gray gradient and apply a steep curve.
but in real world photos, it's really difficult for me to find the difference, even for photos captured with MFDB.
A few comments:
Your upper window is identified as RGB/16 in Photoshop. Mix up?
Secondly, the advantages of 16 bit mainly affect gradations. Here is a dark blend which has been processed using a steep curve:
(http://forums.rennlist.com/upload/8v16bit.jpg)
8-bit on top, 16 bit below. You can see that the 8-bit file produced a rougher transition.
[a href=\"index.php?act=findpost&pid=93236\"][{POST_SNAPBACK}][/a]
-
A few comments:
Your upper window is identified as RGB/16 in Photoshop. Mix up?
Secondly, the advantages of 16 bit mainly affect gradations. Here is a dark blend which has been processed using a steep curve:
(http://forums.rennlist.com/upload/8v16bit.jpg)
8-bit on top, 16 bit below. You can see that the 8-bit file produced a rougher transition.
[a href=\"index.php?act=findpost&pid=93236\"][{POST_SNAPBACK}][/a]
That transition had been generated by only one curve adjustment? Imagine what happens to your 8 bit file when you need to do multiple adjustments to curves, levels or colors.
Yes, you will get away with a few adjustments but when you need to do heavy work on an image an 8 bit file runs the risk of getting completely crushed.
That is basically the criterium I have for choosing 16 or 8 bit. Few adjustments 8, heavy 16 (if only to stay on the safe side).
-
sometimes blue sky colors break in steps with 8bit files when processed even slightly. especially light blue tones are sometimes not rendered fine enough. this are motifs i find often in my work, so i usually work in 16bit. i haold for archival a edited file in full resolution in 16bit, if there will required in any moment changes in the images, for the clients i provide 8bit files in the required image sizes.
upsampling a 8bit filein ps to 16bit and processing it than in 16bit helps till a certain degree to avoid steps in gradations, but it cant invent info in teh image where is no info in the raw data.
i used first time my e75 in compressed mode ( i just took some days of holidays...!!! - and i tried my new contax 645 with the e75 together ).
this results in compressed 16bit files, which has the same image info than 8bit files.
its the same type of compression leica is using with their m8. the info is cutten out more and more as the image data goes up to the higlight levels, where the visual sensitivity of the human perceiption goes down.
i have not noticed bad effects till now, but they should appear theoretically ( and practically ) if the white balance is changed significant or the lights have to be pulled down or up. lets see. for jobs i will go on in 16bit uncompressed.
-
16-bit workflow is well worth it. Here's a practical real-life demo you can try for yourself:
http://www.visual-vacations.com/Photography/16_vs_8.htm (http://www.visual-vacations.com/Photography/16_vs_8.htm)
-
just a quick question regarding the depht of color, that doesn't require a new topic,
can someone tell me if the difference beetwen a 12bit camera and 16bit camera is hardware or software.
Thanks.
-
can someone tell me if the difference beetwen a 12bit camera and 16bit camera is hardware or software.
Hardware, specifically the analog-to-digital converter(s) associated with the sensor.
-
Hardware, specifically the analog-to-digital converter(s) associated with the sensor.
[a href=\"index.php?act=findpost&pid=93295\"][{POST_SNAPBACK}][/a]
Sensor+A/D converter(s)+on-board processor+embedded software (firmware).
-
Thanks, Jonathan, for your demo
but I think that's the case of linear conversion.
CaptureOne's linear conversion appears to be much brighter than other raw converter's. So there are no that much detail compressed into the shadow.
Here is a comparison, a steep curve applied...
i don't see a substantial difference between 16b and 8b version.
16-bit workflow is well worth it. Here's a practical real-life demo you can try for yourself:
http://www.visual-vacations.com/Photography/16_vs_8.htm (http://www.visual-vacations.com/Photography/16_vs_8.htm)
[a href=\"index.php?act=findpost&pid=93247\"][{POST_SNAPBACK}][/a]
-
Can we assume you are converting to 8-bit BEFORE applying the curve? Just want to clarify.
-
Is it is really worth using 16 Bit TIFF? 16b TIFF heavily slows down photoshop, (especially in PC systems, which have 2GB RAM limit for photoshop)
[a href=\"index.php?act=findpost&pid=93229\"][{POST_SNAPBACK}][/a]
I didn't notice anyone mention it, but the rgb colour space being used is relevant to the bit depth. The bigger the gamut, the more potential you have to do damage.
I think its a judgment call depending on the type of images you are handling, but if I decide to use it I usually get out of 16 bit asap after a significant adjustment has been done as generally there is only one big shift for most images. I am using as g5 dual 1.8 with 2 G ram, and it struggles with 16 bit files from 22 MP.
If my Mac was making easy work of it I would think differently, but for now it depends on how many files need doing and how hard I plan to attack an image. I think 16 bit is a bit over-rated for normal processing requirements.
David
-
Can we assume you are converting to 8-bit BEFORE applying the curve? Just want to clarify.
[a href=\"index.php?act=findpost&pid=93435\"][{POST_SNAPBACK}][/a]
Negative, both 16b & 8b version are directly from C1(in AdobeRGB space)
Also tried CameraRAW fro the P45's files...
Also tried Aptus75's files with LeafCapture 10..
results are the same
16b version shows no substantial advantages
-
Negative
If you are not working in 8-bit mode for the 8-bit sample, including applying the curve, then it defeats the purpose of the test.
-
Negative, both 16b & 8b version are directly from C1(in AdobeRGB space)
Also tried CameraRAW fro the P45's files...
Also tried Aptus75's files with LeafCapture 10..
results are the same
16b version shows no substantial advantages
[a href=\"index.php?act=findpost&pid=93461\"][{POST_SNAPBACK}][/a]
The difference won't be seen in the files exported from the RAW processing program, if the curve was applied in that program... but the difference will be visible in the files that have curves applied to them after in Photoshop. Depending on the image you will see more or less of a difference, but in general you can push a file around a lot more within Photoshop if it is in 16 bit mode.
-
for anything thats going to print CMYK, lets remember that the printer is gonna take that file, convert to 8bit and convert to 4color, basically negating a bunch subtle work. I saw it on a daily basis when I worked at a color house....photog's files trashed by people who dont know any better/dont care...
however, I'd rather have the addt'l info while in retouching mode...i can see the benefit
-
I shoot architecture. I shoot 16 bit for exteriors with sky, 'most everything else 8 bit.
-
Should there be a discussion of both the output & editing devices used for images when debating 16 vs. 8 bit?
Canon's new printer has a 16 bit printing mode which takes advantage of 16 bit files to make prints with a greater range of colors.
For current commercial work I deliver 8 bit files to the client, but I edit, retouch, and archive in 16 bit. It may be, in 5 to 10 years, that all the desktop printers and monitors will be using 16 bit information. They may also be able to reproduce an Adobe1998 or larger colorspace.
I would feel defeated if all of my better images were saved in 8 bit sRGB when that time comes around.
-Eric
-
The difference won't be seen in the files exported from the RAW processing program, if the curve was applied in that program... but the difference will be visible in the files that have curves applied to them after in Photoshop. Depending on the image you will see more or less of a difference, but in general you can push a file around a lot more within Photoshop if it is in 16 bit mode.
[a href=\"index.php?act=findpost&pid=93468\"][{POST_SNAPBACK}][/a]
the curve is applied in photoshop, of course
8bit mode for the 8bit image
16bit mode for the 16bit image
-
Can we assume you are converting to 8-bit BEFORE applying the curve? Just want to clarify.
[a href=\"index.php?act=findpost&pid=93435\"][{POST_SNAPBACK}][/a]
sorry, i meant that i didn't convert a 16b image to 8b mode before applying the curve.
-
Thanks, Jonathan, for your demo
but I think that's the case of linear conversion.
CaptureOne's linear conversion appears to be much brighter than other raw converter's. So there are no that much detail compressed into the shadow.
Here is a comparison, a steep curve applied...
i don't see a substantial difference between 16b and 8b version.
That's because you're applying the curve internally in Capture One (which is always in 16-bit mode), and then downsampling to 8-bit after the curve is applied. Your test is not valid. Export the dark image into Photoshop, convert to 8-bit mode, then apply the curve, and you'll get something much different.
-
That's because you're applying the curve internally in Capture One (which is always in 16-bit mode), and then downsampling to 8-bit after the curve is applied. Your test is not valid. Export the dark image into Photoshop, convert to 8-bit mode, then apply the curve, and you'll get something much different.
[a href=\"index.php?act=findpost&pid=93617\"][{POST_SNAPBACK}][/a]
i said, for twice, the curve was applied in photoshop.
the question is:
there is huge difference between 16b and 8b images' histogramafter a very steep curve, the 8b image's histogram did turn very discrete whereas the 16b image's histogram remained continuous.
BUT why there is no such obvious difference when comparing the images, on an Apple Cinema Display HD.
there is nearly no difference in shadow
a little bit more in mid-tone, and finally gets a little bit obvious in hightlight
maybe PhaseOne's lossless compression is in fact not lossless
maybe 16 bit images only depict CCD noise more smoothly than 8 bit ones for this generation of MFDB. (that's why the histogram is smoother but the image is not)
-
I think you are doing something wrong in your workflow. Jonathan's sample makes the difference very clear. 16-bit is very important is you are going to be applying curves.
Try it for yourself.
I just downloaded the file from http://www.visual-vacations.com/Photograph...2-0021_RT16.TIF (http://www.visual-vacations.com/Photography/MiscImages/2004-07-02-0021_RT16.TIF)
Here are my results using the same curve, first in 16 bit then in 8 bit
(http://forums.rennlist.com/upload/16_copy2.jpg)
(http://forums.rennlist.com/upload/8_copy11.jpg)
See the poor quality of the 8-bit version?
-
Perhaps the problem is that your sample image is extremely noisy to begin with, so posterization isn't as noticeable because it's masked by the high noise level. But if you use a clean image, gradient posterization is a lot more noticeable.
-
maybe PhaseOne's lossless compression is in fact not lossless
maybe 16 bit images only depict CCD noise more smoothly than 8 bit ones for this generation of MFDB. (that's why the histogram is smoother but the image is not)
[a href=\"index.php?act=findpost&pid=94049\"][{POST_SNAPBACK}][/a]
If it is showing *sensor* noise more accurately, then it is also showing the image more accurately, because sensor noise *is* signal, and only called noise because it is generally unwanted. If the act digitizing at 16 bits, however, is the cause of more read noise in the more significant bits, then it is counter-productive.