Ad
Ad
Ad
Pages: « 1 [2] 3 »   Bottom of Page
Print
Author Topic: Monochrome r/g/b images from raw?  (Read 7312 times)
Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 7121


WWW
« Reply #20 on: May 03, 2012, 06:56:09 PM »
ReplyReply

1) no demosaicking happened

2) pixels = same type sensels (only R sensels or only B sensels or only G1 sensels or only G2 sensels - other channels are totally removed)

OK, got that, but the software has to be converting that information into something viewable - probably grayscale pixels (for each primary) representing each of those sensels at 16-bit depth. So then two questions arise: if the purpose is to see the grayscale image of each of the R, G, or B primaries, what advantage is there going about it this way, versus Photoshop's disaggregation of the demosaiced data in the Channels palette? Would it tell us much of practical usefulness about the quality of the demosaicing? I'm asking because I'm wondering.
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
dreed
Sr. Member
****
Offline Offline

Posts: 1291


« Reply #21 on: May 03, 2012, 07:03:51 PM »
ReplyReply

Ok, so I tried RawAnalyze and it gets a bit confused when you select only one channel to work with when deciding what is and isn't 100% zoom

OK, got that, but the software has to be converting that information into something viewable - probably grayscale pixels (for each primary) representing each of those sensels at 16-bit depth. So then two questions arise: if the purpose is to see the grayscale image of each of the R, G, or B primaries, what advantage is there going about it this way, versus Photoshop's disaggregation of the demosaiced data in the Channels palette? Would it tell us much of practical usefulness about the quality of the demosaicing? I'm asking because I'm wondering.

I wasn't aware that Photoshop could work with the separate channels of raw files principally because it doesn't know about raw files at all - Adobe Bridge does. Or am I missing something?

Update:
As a case in point regarding the question of "what does it tell us", as can be seen in the earlier attached images, the demosaic'd colour file has fringing where the white of the snow on the mountain meets the blue sky. There is no such artifact when using a single (in this case red) channel from the raw file. At the very least, with no sharpening it suggests that the demosaic is introducing colour artifacts.
« Last Edit: May 03, 2012, 07:06:25 PM by dreed » Logged
Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 7121


WWW
« Reply #22 on: May 03, 2012, 07:07:06 PM »
ReplyReply

I should expand on this a little...

My interest here was to see how B&W images that result from using just a single channel of pixels compares with downsizing a colour image converted into B&W.

Also in this test is to see if there is less chromatic aberration due to different wavelengths focusing at different locations on the sensor.

Finally, it gives me a good sense for how much sharpness I should be applying in LR. Before being able to look at the per-channel images, I was just guessing about how much correction to apply (if any.)

The catch is that the output from 4channels.exe does need work (brightening, etc) before it is really usable.

As an example of the difference, I've included 3 attachments. One is just the red channel from the file, one is colour (no adjustments) from LR, and one is after colour to B&W (no other adjustments) in LR. Of course the catch here is that the red channel is 1/4 of the size of the other two.

Well, on your first point, of course there is no comparison because a conversion to B&W relies on all three primaries, which inherently must be a very different result from that given by any one of the constituent primaries in islolation, so not clear what's the purpose behind it.

On the second point, CA is the result of what your lens does when the image is exposed and LR has profiles and manual controls for correcting it very efficiently, so again, I may be missing something but I have trouble perceiving the value-added of the difficult approach.

On the third point, there is excellent instruction available for LR4 on how to evaluate and control sharpness using the grayscale rendiition of the image that you can trigger easily when using the sharpening panel tools.

So I remain somewhat mystified about the usefulness of bothering with the hard way, but of course an interesting intellectual odyssey from which something useful could emerge - so I'm not knocking it, just wondering about value-added and usefulness.
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
deejjjaaaa
Sr. Member
****
Offline Offline

Posts: 1124



« Reply #23 on: May 03, 2012, 07:07:34 PM »
ReplyReply

OK, got that, but the software has to be converting that information into something viewable - probably grayscale pixels (for each primary) representing each of those sensels at 16-bit depth.

yes, certainly when you visualize it to see w/ your eyes (and not to feed into some software to crunch that numbers) then you (software) do some operations to ultimately display the data... but you do some operations to display the data from jpgs for example... raw file is an image, the mere fact that WB is not applied, curves are not applied, demosaick did not happen, data is not transformed into some color space, etc does not make raw file a non image... you can view it with a proper software like you view your jpgs... you can view all 4 channels at once or you can view it per channel... get rawdigger - it is a nice visualizer.



Logged
deejjjaaaa
Sr. Member
****
Offline Offline

Posts: 1124



« Reply #24 on: May 03, 2012, 07:09:10 PM »
ReplyReply

Ok, so I tried RawAnalyze

rawanalyze is not updated anymore as its author, mr Gabor Schorr, died in 2010... use rawdigger.
Logged
Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 7121


WWW
« Reply #25 on: May 03, 2012, 07:10:42 PM »
ReplyReply

Ok, so I tried RawAnalyze and it gets a bit confused when you select only one channel to work with when deciding what is and isn't 100% zoom

I wasn't aware that Photoshop could work with the separate channels of raw files principally because it doesn't know about raw files at all - Adobe Bridge does. Or am I missing something?

Update:
As a case in point regarding the question of "what does it tell us", as can be seen in the earlier attached images, the demosaic'd colour file has fringing where the white of the snow on the mountain meets the blue sky. There is no such artifact when using a single (in this case red) channel from the raw file. At the very least, with no sharpening it suggests that the demosaic is introducing colour artifacts.

Photoshop isn't working with separate channels of raw files. It is working with separate channels of rendered raw files - different things.

What I see in your second image is a pure case of CA - it most likely has nothing to do with demosaicing, and you would of course not see it through one channel alone - that doesn't mean it isn't there and that it wasn't created by your lens.
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 7121


WWW
« Reply #26 on: May 03, 2012, 07:14:48 PM »
ReplyReply

rawanalyze is not updated anymore as its author, mr Gabor Schorr, died in 2010... use rawdigger.

That's really sad to learn. Gabor was a respected, expert contributor to the digital imaging community.
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 7121


WWW
« Reply #27 on: May 03, 2012, 07:18:37 PM »
ReplyReply

yes, certainly when you visualize it to see w/ your eyes (and not to feed into some software to crunch that numbers) then you (software) do some operations to ultimately display the data... but you do some operations to display the data from jpgs for example... raw file is an image, the mere fact that WB is not applied, curves are not applied, demosaick did not happen, data is not transformed into some color space, etc does not make raw file a non image... you can view it with a proper software like you view your jpgs... you can view all 4 channels at once or you can view it per channel... get rawdigger - it is a nice visualizer.

OK, I hear you - nice for visualizing data that has been tinkered with as little as possible while still allowing one to see what the sensor captured. Not a bad idea, and even better if I could conjure-up what I would use it for, apart from the intellectual curiosity.
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
dreed
Sr. Member
****
Offline Offline

Posts: 1291


« Reply #28 on: May 03, 2012, 08:03:27 PM »
ReplyReply

I wonder what impact (if any) it has on noise, as rather than deal with noise from all three colour channels, there is only one.

Does that make noise easier to manage, harder, less or more prevalent?

Or is there no difference at all?
Logged
Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 7121


WWW
« Reply #29 on: May 03, 2012, 08:12:45 PM »
ReplyReply

There is good noise reduction software on the market allowing you to manage noise channel by channel.
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
jerryrock
Sr. Member
****
Offline Offline

Posts: 570



WWW
« Reply #30 on: May 03, 2012, 11:50:42 PM »
ReplyReply

There is good noise reduction software on the market allowing you to manage noise channel by channel.

Photoshop
Logged

Gerald J Skrocki
skrockidesign.com
dreed
Sr. Member
****
Offline Offline

Posts: 1291


« Reply #31 on: May 04, 2012, 12:53:59 AM »
ReplyReply

Well then maybe being able to do this remains something just for those that are curious Smiley
Logged
Mark D Segal
Contributor
Sr. Member
*
Offline Offline

Posts: 7121


WWW
« Reply #32 on: May 04, 2012, 05:21:36 AM »
ReplyReply

Well then maybe being able to do this remains something just for those that are curious Smiley

I agree, and I would say not only that, but perhaps useful to software developers, programmers, code writers for any analytic insight it can provide for their work.
Logged

Mark D Segal (formerly MarkDS)
Author: "Scanning Workflows with SilverFast 8....." http://www.luminous-landscape.com/reviews/film/scanning_workflows_with_silverfast_8.shtml
PierreVandevenne
Sr. Member
****
Offline Offline

Posts: 512


WWW
« Reply #33 on: May 04, 2012, 05:52:55 AM »
ReplyReply

Iris

use either the menus or the splitCFA command

http://www.astrosurf.com/buil/iris/tutorial5/doc17_us.htm
Logged
dreed
Sr. Member
****
Offline Offline

Posts: 1291


« Reply #34 on: May 10, 2012, 05:25:00 AM »
ReplyReply

Does colour moire exist in a single channel of an image captured with a Bayer matrix sensor?

If so, what does it look like in B&W?
Logged
FranciscoDisilvestro
Sr. Member
****
Offline Offline

Posts: 569


WWW
« Reply #35 on: May 10, 2012, 06:28:32 AM »
ReplyReply

There might be aliasing or moire, but no "colour". It will look as a pattern not present in the original captured scene.

You will actually get more aliasing than if you use the 4 channels and perform demosaicing. When using only one raw channel, you have 1/4 of the resolution, and even if your camera had an AA filter, it would be too weak for just one channel.
Logged

dreed
Sr. Member
****
Offline Offline

Posts: 1291


« Reply #36 on: May 10, 2012, 07:42:48 AM »
ReplyReply

There might be aliasing or moire, but no "colour". It will look as a pattern not present in the original captured scene.

You will actually get more aliasing than if you use the 4 channels and perform demosaicing. When using only one raw channel, you have 1/4 of the resolution, and even if your camera had an AA filter, it would be too weak for just one channel.

Well, I for one would like to see if and how it is different when just a single channel is looked at.

Note, this is only applicable when moire is visible at 100% when rendered raw files.
Logged
FranciscoDisilvestro
Sr. Member
****
Offline Offline

Posts: 569


WWW
« Reply #37 on: May 10, 2012, 09:16:04 AM »
ReplyReply

Some interesting info in this thread:

http://www.luminous-landscape.com/forum/index.php?topic=58661.0
Logged

dreed
Sr. Member
****
Offline Offline

Posts: 1291


« Reply #38 on: May 10, 2012, 04:07:53 PM »
ReplyReply


All of the images in that thread were created by rendering an image from the raw file, they were not created by just extracting a single channel's data.

And as a "for example" of why I find this interesting, Leica's new 18MP monochrome camera is without an AA filter...
Logged
FranciscoDisilvestro
Sr. Member
****
Offline Offline

Posts: 569


WWW
« Reply #39 on: May 10, 2012, 07:34:44 PM »
ReplyReply

And as a "for example" of why I find this interesting, Leica's new 18MP monochrome camera is without an AA filter...


It is still possible to get images with aliasing even if you don't perform interpolation, so the M Monochrom will not be inmune to this.
Aliasing happens when the frequency of the signal is above Nyquist frequency of the sampling device.


Logged

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

Ad
Ad
Ad