Four raw converters , handling demosaic errors

[David Eichler]

Two thoughts:
- Defaults are meaningless. How do the files handle under each processor when you're trying to get the best from each?
- Look at those smoooooth diagonal lines on the C1 version and see how none of the others come close to rendering those diagonal lines with such smoothness? This is the sort of thing that matters a LOT when you go to print big, especially if you need to interpolate upward to get there.

Sorry, not seeing what you mean about the smoothness of those lines relative to the others, except for the bottom right one, which appears to have some digital "sharpening" applied relative to the other three.

[David Eichler]

Hi,

I converted an image from my P45+ using four different raw converters. The image was chosen as it had a lot of aliasing artefacts. What I wanted to see was if any of these raw converters handled the problem areas well.

The raw images is here: http://echophoto.dnsalias.net/ekr/Articles/RawConverters/Raws/CF045470.IIQ

The two attached crops are taken from different points of the image. They have been resized in Photoshop to 200% using bicubic interpolation for comfortable viewing. I tried to process the images as close to default as reasonable, but turned of sharpening off.

What I see is that Lightroom conversion is pretty bad on the crane part, aside from colour aliasing it also has some halo effect that is not present in the images. Sharpening was here set to zero, so it is not a sharpening artefact.

RawTherapee	CaptureOne 7.3
Lightroom 5.6	AccuRaw 2.2.6

The second crop is a more normal part of the image.

These conversions were not intended to be artful, the intention was to all raw developers near deafults. I tried to turn sharpening off in all developers.

Best regards
Erik

What is going on with the color artifacts I see in the tower to the left of the crane? And, the one that shows it least, on the bottom right, also appears to have some sharpening applied relative to the others, which makes me wonder if that version also has more basic processing applied than the others.

As someone mentioned, "defaults" might not be the best place to compare, since each software might have a different starting point and different criteria for
default settings. I would like to see comparisons of the best you think you can get out of each software with a particular image, although the best test would be to compare with a variety of images, imo.

[ErikKaffehr]

Hi,

There is now sharpening on any of the images, AFIK. AccuRaw, the one on the bottom right is supposed to extract more detail.

Best regards
Erik

What is going on with the color artifacts I see in the tower to the left of the crane? And, the one that shows it least, on the bottom right, also appears to have some sharpening applied relative to the others, which makes me wonder if that version also has more basic processing applied than the others.

As someone mentioned, "defaults" might not be the best place to compare, since each software might have a different starting point and different criteria for
default settings. I would like to see comparisons of the best you think you can get out of each software with a particular image, although the best test would be to compare with a variety of images, imo.

[synn]

Erik, the point is, if a software assigns, let's say  "50" as a sharpening value, it does so for a good reason. pushing it to zero is not the best way to showcase its capabilities. Another software that shows "0" as default may already be applying more than "50" without letting you know.

If you're going for "Default conversions", let every slider be at default. Even sharpness. Else, process them the best you can.

[ErikKaffehr]

Hi Synn,

As you see, there is a massive amount of aliasing of every kind on the cranes and also on some of the roofs. What I want to find out which raw converter produces the image with least artefacts and most natural detail.

These converters are different, in the sense that Lightroom and Capture One can actually process an image to be ready for printing. RawTherapee and AccuRaw are just only raw converters, produce images that need to be processed in Photoshop.

The idea of not using sharpening is that better sharpening methods, like those used by Bart, would be applied and I don't want sharpening artefacts, which in the next processing step would be enhanced.

Best regards
Erik

Erik, let me have a go at the file later today. Will post the results up later.
I don't understand though, that you insist on keeping everything at default, but push sharpening to zero. All bayer files are assigned a default sharpening level by editing softwares upon import and for good reason. pushing them downwards to zero is not a reflection of any real world outcome.

If I were to do this test, I would edit all files to the best of what the softwares can achieve. doing four default renders is akin to driving a bunch of supercars down a city street to assess what their engines are capable of.

[Schewe]

You mean to say that software written by John Knoll and sold millions of times is not good enough for you when it is good enough for everybody else?

Not for nothing, but John had ZERO to do with ACR/LR...it was his brother Thomas :~)

[synn]

I agree with you. Lightroom is not really a contender, C1 is ok, but AccuRaw seems to do best, at least on this image.
Frankly, I would have expected C1 to go the extra mile in sharpness on a Phase file, but since a few versions down the line my impression is it has become slightly more blurry. I ran a lot of P45+ files through it too, and for Phase files it consistently beat Lightroom in color and detail.

Edmund

There's either more than one person posting from your account or you're actively trolling these forums for a reaction; because this post is completely contradicting what you were saying a few hours ago.

Either way, there's no one else on the board that has so many posts under their belt without contributing anything worthwhile OR showing any of their work. Carry on dancing...

[synn]

Hi Synn,

As you see, there is a massive amount of aliasing of every kind on the cranes and also on some of the roofs. What I want to find out which raw converter produces the image with least artefacts and most natural detail.

These converters are different, in the sense that Lightroom and Capture One can actually process an image to be ready for printing. RawTherapee and AccuRaw are just only raw converters, produce images that need to be processed in Photoshop.

The idea of not using sharpening is that better sharpening methods, like those used by Bart, would be applied and I don't want sharpening artefacts, which in the next processing step would be enhanced.

Best regards
Erik

I am not a sharpening guru or anything, but my understanding is that sharpening happens in three stages. RAW presharpening, proper sharpening done in PS/ Focus Magic etc. and finally, output sharpening, depending on your print format.

There's nothing wrong with applying a bit of presharpening in your RAW converter. In fact, there's a lot to gain from it. In all my time processing RAW files, I have never seen any sharpening artifacts that arose as a result of some moderate presharpening.

[yaya]

As a general note, turning sharpening off in meaningless without adjusting noise reduction accordingly (and clarity, if the converter offers such facilities).

And Mr. Knoll's name is Thomas, not John...

[EricWHiss]

Defaults aren't standardized, and I agree with Synn.  I can't imagine getting a meaningful comparison without someone who really knows how to work a particular software for RAW conversion for a particular camera, i.e. Schewe + LR/ACR against Doug P + C1.      There's also a lot more facets to software than how it handles color aliasing and one who focuses solely on that probably is missing something.  

[torger]

Testing a particular aspect of a raw converter can be interesting, and focusing on one aspect alone makes it easier to compare. Of course you would not base the choice on one aspect, but it's still interesting.

For MF color aliasing is particularly interesting as the sensors don't have antialias filters. There can also be a conflict between extracting detail and avoiding aliasing, eg the best detail extractor might not be the best choice for a sensor without AA filter.

In RawTherapee you can choose between several different demosaicers by the way, but Emil Martinec's Amaze is the default, and best for all-around use. If you need more robustness VNG4 is better, but then you loose a little detail. There's also a couple of demosaicers adapted for noisy images (high ISO). I would guess some of the commercial raw converters also have more than one demosaicer, but change them automatically depending on raw file and other settings rather than manually like in RT.

If you only look at aliasing you don't need to worry much about minor noise reduction or sharpening settings, it won't significantly alter the false color aliasing, as it's an error occurring at the demosaicing level. If those settings would change demosaicing algorithm as a side effect (not likely) it will have effect though of course. How "jagged" a diagonal line will look (which is also a form of aliasing) can however be affected by sharpening etc.

[Bart_van_der_Wolf]

Erik, the point is, if a software assigns, let's say  "50" as a sharpening value, it does so for a good reason. pushing it to zero is not the best way to showcase its capabilities. Another software that shows "0" as default may already be applying more than "50" without letting you know.

Hi Sandeep,

While it is true that all Raw conversions need some sort of sharpening, the defaults are not necessarily (or rather they rarely are) optimal. But that's also not what Erik is trying to figure out. Adding sharpening, will only exaggerate artifacts that are already present in the baseline conversion.

So, Erik is looking for a well-behaved baseline Raw conversion, on which to base further processing. The trouble is that a Bayer CFA sensor without optical low-pass filter (OLPF) will pose any Raw converter for an almost impossible task, because the source data is compromised (with undersampled and aliased data). With the 'right' kind of subject matter, things will become even more challenging.

I think that the visibility of False Color aliasing artifacts can often be mitigated by using the proper correction tools (luminance aliasing is almost impossible to correct without elaborate post-conversion editing). It seems logical that converters such as Capture One, which natively are designed to deal with non-AA filtered images, would be able to tackle the issue better than others, such as e.g. LR/ACR, that are more used for conversion of low-pass filtered (or shot with lesser optics) Raw files.

If you're going for "Default conversions", let every slider be at default. Even sharpness.

I disagree. Default sharpening settings, especially those of LR/ACR (complicated with noise reduction being linked to the same sharpening settings), are more often wrong than correct. They are set to produce acceptable results for casual use, but they act oblivious about actual lens performance, and cannot have a clue as to what the intended output will be. Down-sampling or upsampling of the conversion result will require different settings, unless one only applies Capture sharpening (and frankly the converters offer virtually no guidance for that at all).

Else, process them the best you can.

Yes, but that's not the current issue that Erik is addressing, it's step number two. Besides, each program will produce somewhat different quality of results, and they offer different workflows, so that may also cloud the comparison and make it more subject matter dependent. Each program also has a learning curve, so it would take considerable experience/time to get the most out of each one of them.

Cheers,
Bart

[ErikKaffehr]

Hi,

I did turn off noise reduction, I guess I need to elaborate on using defaults.

Thanks for pointing out it's Thomas Knoll, I felt it was the wrong Knoll, but I mixed up

Best regards
Erik

As a general note, turning sharpening off in meaningless without adjusting noise reduction accordingly (and clarity, if the converter offers such facilities).

And Mr. Knoll's name is Thomas, not John...

[eronald]

As a general note, turning sharpening off in meaningless without adjusting noise reduction accordingly (and clarity, if the converter offers such facilities).

And Mr. Knoll's name is Thomas, not John...

I apologize for getting the name wrong - senior moment - my bad.

As regards color, obviously C1 can and indeed is better than any other software on P1 backs, as P1 have complete information on their sensor and cover, including spectral sensitivity, ample time to test it, and time to design some very good profiles for it. When I owned their back I did some testing, in fact wrote my own profiler and profile edit suite, and my feeling has always been that both the linearisation/calibration of the Phase backs themselves and the various profiles made by Phase for various shooting conditions with those backs are exemplary, but this does not apply to third party cameras for which they haven nowhere near the same level of preparation or motivation.

Lightroom color is a complicated story, seen from the outside; at the beginning their camera measurements were not so good -spiky lightbox -  which accounted for some of their issues. I am sure their camera measurements have caught up now, but in the last few years they have decided to do some strange "hue twisting" where hue is dependent on luminance, which has been documented. I gave up on LR color long ago, and use the native converter of whoever made the camera I use.

When it comes to sharpness, I think the only way, as other posters have suggested is to get experts for each converter to do their own cooking and compare the dishes ...


Edmund

[Paul2660]

Eric

I also noticed the colors on the tall antennae left of the crane.  I wasn't sure it that was all color aliasing or if the antennae was painted to some of the colors.

Thanks
Paul

[Manoli]

...  profiles made by Phase for various shooting conditions with those backs are exemplary, but this does not apply to third party cameras for which they haven nowhere near the same level of preparation or motivation.

Edmund,

That may have been true for V7, it may still be true for some non-Phase cams. But in V8, I see a distinct improvement in the demosaic-ing of both Fuji x-trans and Sony A7 files, as well as pleasing colour rendition.

M

[eronald]

Edmund,

That may have been true for V7, it may still be true for some non-Phase cams. But in V8, I see a distinct improvement in the demosaic-ing of both Fuji x-trans and Sony A7 files, as well as pleasing colour rendition.

M

Manoli,

 I'm saying that when C1 is used on *their own* files their color is usually wonderful. On third party cameras ... it varies. Some I've seen were good, some nothing special. I'm glad to hear your Fuji and Sony results are good.

C1 used to work well with camera profiling. In fact, many years ago I sold some edited profiles for C1, and some of the profiles I made really improved the renderings, others were epic failures; in particular I remember the Canon 5D which was a very nice camera, but for some reason the profiles I made worked well with the camera samples I had and failed with others - maybe there were several sensor suppliers.  

 One interesting and underestimated converter is Canon's DPP which has got good color and exceptionally good sharpening probably because Canon know a lot about their own lenses and sensors, and implemented some form of deconvolution.

Edmund

[Paul2660]

Manoli,

 I'm saying that when C1 is used on *their own* files their color is usually wonderful. On third party cameras ... it varies. Some I've seen were good, some nothing special. I'm glad to hear your Fuji and Sony results are good. Many years ago I sold some profiles for C1, and some of the profiles I made really improved the renderings, and some of my own profiles for other cameras were epic failures; in particular I remember the Canon 5D which was a very nice camera, but for some reason the profiles I made worked well with the camera samples I had and failed with others - maybe there were several sensor suppliers. 

 One interesting and underestimated converter is Canon's DPP which has got good color and exceptionally good sharpening probably because Canon know a lot about their own lenses and sensors, and implemented some form of deconvolution.

Edmund

I would tend to agree with this.  C1 is great on the Phase/Leaf backs, but can be hit and miss on other cameras.  It seems very good for example on Fuji X-trans, but IMO C1 never got the D800/e right.  Profile or what ever was needed, where as the use of the IQ250 profile can often create a better overall image.  D1 did seem to do a better job on the D810.

Canon's converter is excellent, however it's tool set is rather basic and outdated.  It can be a great starting point for some files, but I can't remember if it will output to a DNG, as I recall it won't. 

The real power to me of LR and C1 is in their toolsets.  I just never seem to need much in CC anymore, unless it's a bit of a Topaz plugin or Focus Magic.  C1 8 with the additions they made to the local adjustments is now very powerful and an excellent tool, (if it only had a history tab like LR) since the more you do, the harder it is to get back a step or two especially if you are working several images at once. 

Paul

[eronald]

I would tend to agree with this.  C1 is great on the Phase/Leaf backs, but can be hit and miss on other cameras.  It seems very good for example on Fuji X-trans, but IMO C1 never got the D800/e right.  Profile or what ever was needed, where as the use of the IQ250 profile can often create a better overall image.  D1 did seem to do a better job on the D810.

Canon's converter is excellent, however it's tool set is rather basic and outdated.  It can be a great starting point for some files, but I can't remember if it will output to a DNG, as I recall it won't.  

I don't think Nikon or Canon give away all the secrets of their in-camera Raw processing, or even the white balance info and illuminant information

To a point, Nikon, Sony and Canon are competing with Phase so the C1 software is written with imperfect information when it comes to third party files. And illuminant details may be much more pertinent than greater profile accuracy ...and cameras contain other sensors in addition to the big image sensor

Phase one excel at calibrating their backs, and they supply superb profiles for standard situations such as daylight or matched studio flash.This is where pros work, and luckily where the backs are at their best. But when a Nikon or Canon camera is used by a consumer in mixed light, illuminant information and scene information will make all the difference, and this is probably stuff which the manufacturers do not wish to show their competitors if they can help it.

So, I would expect DPP or Nikon Capture to win in color with default settings every time a camera is used in mixed light, not because the converter is intrinsically better or the profiles are any better,  but because it knows more about the camera and conditions under which the image is taken.

Horses for courses?

Edmund

[ErikKaffehr]

Hi Paul,

I am pretty sure it is aliasing. I will try to reshoot the scene in a week or two, as I want to find out if some of the artefacts I see from Lightroom are P45+ only or also affecting OLP filtered images. So the plan is to shoot both DSLR and P45 and also make a telephoto shot to have a reference image. But I can only shoot holidays as the days are now short.

Best regards
Erik

Eric

I also noticed the colors on the tall antennae left of the crane.  I wasn't sure it that was all color aliasing or if the antennae was painted to some of the colors.

Thanks
Paul