Demosaicing: a possibly silly and definitely basic question about raw processing

[Ellis Vener]

For raw processing, not camera produced JPEGs.

Are there differences in the demosaicing algorithms used by various raw processing programs (Adobe Camera Raw, Apple Aperture/iPhoto, Canon's DPP, Capture One, DxO, Hasselblad Phocus, Iridient Developer, Nikon's Capture NX programs,  and the similar Sigma and Sony applications, etc.) or is the demosaicing color interpolation done or set  by a camera's internal image processor as part of the raw file?

[Schewe]

Raw means that the image must be demosaiced by software after the fact, and yes, there are differences–sometimes huge–between different software's rendering (except for ACR/LR which is the same assuming the same dot version).

[Ellis Vener]

Raw means that the image must be demosaiced by software after the fact, and yes, there are differences–sometimes huge–between different software's rendering (except for ACR/LR which is the same assuming the same dot version).

Thank you Jeff.

[Ellis Vener]

Raw means that the image must be demosaiced by software after the fact, and yes, there are differences–sometimes huge–between different software's rendering (except for ACR/LR which is the same assuming the same dot version).

A follow up question with regard to ACR, are the different "dot versions" the same thing as different "process" versions (2012, etc.)?

[Schewe]

A follow up question with regard to ACR, are the different "dot versions" the same thing as different "process" versions (2012, etc.)?

No...not necessarily. Some dot versions in the past have altered or changed in a dot version although that should have no impact on demosiacing...the big change in demosiacing happened in ACR 4.1 where the built -in luminance noise reduction was removed.

Process versions can change the way ACR handles the color and tone mapping but not demosiacing.

[Pictus]

Would be very good if a future version of ACR/Lightroom to be accelerated by OpenCL.
Capture One 7 with OpenCL(need capable graphic card) is insane fast.
Also an option to choose a different demosaicing algorithm would be a blessing!
Keep the current as default and  include one for high quality(AMaZE ?)
and another(LMMSE ?) for high ISO and/or moiré...  

Raw file http://movies.dpreview.com.s3.amazonaws.com/pentax_k5iis/IMGP0101.DNG.zip

[Schewe]

Would be very good if a future version of ACR/Lightroom to be accelerated by OpenCL.

Likely at some point...

Also an option to choose a different demosaicing algorithm would be a blessing!

Very unlikely...although don't be surprised that Adobe (read the ACR engineering elves) come up with improved demosaicing algorithms in the future. Not sure when though.

[Tim Lookingbill]

Always found this site on demosaicing algorithm rendering differences interesting...

http://www.linuxphoto.org/html/test_demosaicing.html

Other camera influences mainly regarding hardware as in the presents and thickness of a sensor's antialiasing filter can challenge certain demosaicing approaches.

[EduPerez]

RawTherapee lets the user choose between different demosaicing algorithms.

[Ellis Vener]

Always found this site on demosaicing algorithm rendering differences interesting...

http://www.linuxphoto.org/html/test_demosaicing.html

Other camera influences mainly regarding hardware as in the presents and thickness of a sensor's antialiasing filter can challenge certain demosaicing approaches.

Tim,
That looks great and thank you for the link, but unless I am missing something the tests do not appear to have been updated in awhile.

[Vladimirovich]

Raw means that the image must be demosaiced

but then there is Sigma where you might not need any demosaicking (but you do need w/ recent sensor) - just the color transform... or monochrome cameras - no demosaick either

[Vladimirovich]

Tim,
That looks great and thank you for the link, but unless I am missing something the tests do not appear to have been updated in awhile.

as mentioned Rawtherapee allows to select from many options and it is free - so you can run the tests yourself on your raws

[sniper]

as mentioned Rawtherapee allows to select from many options and it is free - so you can run the tests yourself on your raws

I'm fairly new to Rawtherapee, could you point me to the options to change the demosaicing, I can't see them in the settings.
Regards Wayne

[EduPerez]

I'm fairly new to Rawtherapee, could you point me to the options to change the demosaicing, I can't see them in the settings.
Regards Wayne

It is not part of the global settings, but the processing profile applied to each image: open a RAW file, and go to the last tab (the one that locks like a checkered board) on the right pane.

[sniper]

It is not part of the global settings, but the processing profile applied to each image: open a RAW file, and go to the last tab (the one that locks like a checkered board) on the right pane.

Got it now, many thanks for that.    Now I'll have to have a play.   

[Tim Lookingbill]

Tim,
That looks great and thank you for the link, but unless I am missing something the tests do not appear to have been updated in awhile.

Not sure I understand your point, Ellis.

I posted it as an FYI for those wondering how much demosaicing algorithms of various recipes good or bad can affect Raw previews. I didn't intend it to be considered as empirical evidence of one DA's superiority over another.

If your point was concern over whether the link was relevant to this thread, I'm only going by your thread title.

[Ellis Vener]

Tim,
I'm sorry,. I must have written my response badly. I am truly grateful for your taking the time the time to respond and share that link.  My only concern is that the results posted there do not take into account any changes that might have been made to the various programs between the time that survey was created and now. Specifically with reference to Lightroom (ACR) and Capture One.

[Fine_Art]

The screenshot crop below shows the de-bayering methods of Raw Therapee. The methods used by most software are there.

My understanding is that ACR/Lightroom uses the AHD method with some modification.

If you want to see how AHD would work with the deconvolution in RT or the NR of RT you can test it for free. You can also test how AHD compares to AMAZE. AMAZE wins hands down IMO.

[Tim Lookingbill]

Tim,
I'm sorry,. I must have written my response badly. I am truly grateful for your taking the time the time to respond and share that link.  My only concern is that the results posted there do not take into account any changes that might have been made to the various programs between the time that survey was created and now. Specifically with reference to Lightroom (ACR) and Capture One.

I agree, Ellis.

But that linked site shows what type of artifacts to look for in any number of Raw converters that could indicate causality by either or both the influence by the sensor's AA filter and/or demosaicing algorithm. Some artifacts can't be fixed but at least that site shows the cause and what to look for.

For instance my Pentax K100D's AA filter is quite thin and it explains why some shots taken at certain distances combined with my 6MP sensor produce overly sawtoothed edge artifacts (think low rez vector graphic sharpness) in ACR/LR 100% previews while other shots don't. On some shots I have to dial back the sharpening to get rid of this. Other Raw converters I've used produce an artificial softening with noisy edges (almost like jpeg compression) that I don't particularly like. Some shots of screen windows from a distance show moire while others don't.

Is that the fault of the demosaicing algorithm? And is a test based on one image going to prove it? I don't buy it.

I find tests that use a single image to show better or worse demosaicing results pretty misleading because of the other influences pertaining to distance, resolution and default sharpening. That site just shows what the results look like but I admit doesn't tell the whole story, but neither do discussions of this sort as well.