Monochrome fine art on Sigma Cameras versus high res Bayer conversion?

Started 8 months ago | Discussions
TN Args
TN Args Veteran Member • Posts: 8,434
Re: Monochrome fine art on Sigma Cameras versus high res Bayer conversion?

saltydogstudios wrote:

Shawn67 wrote:

saltydogstudios wrote:

I think maybe 3 times? In one I converted most of the photos to B&W and the other I hated the colors so much that I started using filters in Photoshop for "creative" color effects.

That is interesting, I haven't seen that. Colors are malleable depending upon processing options.

I tend to shoot Sigmas at smaller apertures for maximum resolution.

I tend to shoot full-frame cameras at wider apertures for shallow depth of field.

DOF is a huge part of this. I have shots that are that sharp/good contrast, then the next one looks blurry due to slight changes in focus. Or even subject movement if the flash sync speed is lower than 1/125 or so. See below for an example....

I have 24 megapixel CFA cameras.

The Sony A7 that has an AA filter, and I only have manual focus lenses for it which I tend to shoot wide open.

The Fuji X-Pro2 which has an X-trans sensor, no AA filter, and I have the excellent Fuji lenses for it.

At the bottom of this post I have a photo from the Fuji X-Pro2

https://medium.com/ice-cream-geometry/an-ode-to-the-sigma-merrill-506dd0864169

It approaches the Merrill for microcontrast.

It is not as good as either for Sharpness.

If you're saying you get an image that's "as good as the ones I posted" from my Sigma DP3 Merrill and 42 megapixel Sony A7R2 with your 24 megapixel Bayer you're forgetting one important fact.

Those are extreme crops.

I'm not sure if this is the exact photo I used but it's probably within a few frames.

The Sony was similarly cropped.

Sigma DP3M - full (left), crop (right)

Can at 24 megapixel Bayer crop to this level of detail from that far out? Again having shot 24 megapixel CFA cameras I don't think so.

Fuji X Pro 2 (55-200) from a less tight shot than yours above...

At 100% view...

From a 24mp Leica M240 (Konica M-Hexanon 90mm)

At 100%

There are differences between them but to claim there is some sort of serious blur over the picture is a stretch.

And the example of DOF killing this. Again from the Leica... a shot a couple of seconds after the one above....

and at 100%

Pixel peepers would be quick to say this is due to having a CFA but that isn't the reason. Nose is sharp but not enough DOF for the eyes. Think this was f5.6 or f8 on a 90mm.

BTW, I just converted these to B&W for comparison. I far prefer them in color but wanted to stay consistent with your example.

Shawn

Hey-

Thank you for engaging in a smart, thoughtful way & for posting photo samples. I really appreciate that.

Your images are indeed sharp - both in gestalt and zoomed in. I also like the way you mix strobe & ambient, something I've done a lot of in the past, though now I prefer to do either fully one or the other.

In return - here are two photos.

One shot with a Sigma DP3 Quattro.

The other shot with Fuji X-Pro2 with 80mm f/2.8 lens - regarded as one of Fuji's sharpest lenses.

Both shot in "Standard" color mode (whatever that means for each) and at f/8 and lowest ISO (so ISO 100 on the Sigma, "L" on the Fuji).

I shot them with approximately the same framing & with the exact same studio strobe. I set the white balance in each manually & the exposure in each manually as well with my usual method.

Two photos from each in case focus was off on one. Both shot with a studio strobe, so motion blur was not an issue.

SOOC JPGs and RAW files here:

https://www.dropbox.com/sh/bvyru0xwo887sva/AACn24BfB2btTApWaKtXcKV6a?dl=0

This dress form is often my test subject when I'm trying out different lighting concepts. The tight grain of the fabric & relatively neutral color makes it a very good test subject for "sharpness" as well.

Fuji X-Pro2 w 80mm f/2.8 lens, SOOC.

Sigma DP3, SOOC

I think the Fuji is 24 megapixels and the Sigma is 19.

Sigma top, Fuji bottom.

If you look really close, there's a bit of red lint just after the first seam - sort of opposite where the Acme sign is. I've circled it in black.

In the Sigma this bit of red lent is obvious. In the Fuji, you can barely see it. It may be because of the X-Trans sensor and the relative lack of red pixels - something just a few pixels across that's red may not register as well on the X-Trans sensor.

In general, the Fuji feels soft - this is (I suspect) as much a relic of Fuji's X-trans sharpening algorithms as it is the result of the sensor itself.

The "Patented and Manufactured By" text and the address are much more readable on the Sigma than on the Fuji.

On the non-pixel-peeping level, the Fuji feels sharp - on the pixel peeping level, it almost feels out of focus. Where as the Sigma feels razor sharp - and again both are shot at f/8 and with studio strobes. I took two photos with each just to be sure.

Just to be doubly sure, I took two extra photos with the Fuji - one with the same framing (again just in case the focus was off) and one close up.

Fuji X-Pro2 w 80mm f/2.8 at f/8.

Crop of above Fuji Image.

With this framing, the Acme sign is perfectly readable & the little bit of red lint is readily apparent - but this is nearly 2x as many pixels across as the previous image. (because I was physically closer / had a tighter framing.)

And the fabric still feels blurry, like it's out of focus.

I've blogged about this before - I suspect Fuji's sharpening algorithms are more about "edge detection" and trying to find shapes and Sigma's sharpening algorithms are about trying to find the differences between two pixels. I call this the X-Trans version of "Bayer Blur."

So it would make sense that in something that's all texture (and no "shapes" to grab on to), Sigma would do better than Fuji.

Perhaps with a different demosiacing algorithm in Raw Therapee the Fuji will do better.

In a side by side comparison of the 19 megapixel APS-C Sigma Quattro and the 24 megapixel Fuji X-Trans - for detail, I'm going to have to give it to the Sigma Quattro.

This isn't to say that a 24 megapixel CFA image can't be sharp - a 24 megapixel CFA image can be incredibly sharp - as you've amply demonstrated.

But I don't think a 24 megapixel CFA can compare to a 42 megapixel CFA (the Sony A7R2) or a 19 megapixel Foveon.

Thanks for the interesting analysis. One point I would like to make: you used f/8.

For the Fuji that might just be diffraction limited, with an Airey Disc 2.67 pixels wide. Conventionally 2.5 is the point of onset of diffraction limits.

For the Sigma dp3Q it is 2.4 pixels wide, but IMHO that is even more likely to be diffraction limited, since there is no matrix-related alias blur with Foveon, so you would think the effect might start earlier. I have never seen an analysis on this point, though.

Shooting with f/4 and a good lens makes a lot of sense for ultimate detail comparisons, IMHO.

cheers

 TN Args's gear list:TN Args's gear list
Sigma dp0 Quattro Panasonic Lumix DMC-GX7 Olympus E-M5 II Olympus M.Zuiko Digital ED 9-18mm F4.0-5.6 Panasonic Lumix G 14mm F2.5 ASPH +7 more
saltydogstudios Senior Member • Posts: 1,038
Re: Monochrome fine art on Sigma Cameras versus high res Bayer conversion?

TN Args wrote:

Thanks for the interesting analysis. One point I would like to make: you used f/8.

For the Fuji that might just be diffraction limited, with an Airey Disc 2.67 pixels wide. Conventionally 2.5 is the point of onset of diffraction limits.

For the Sigma dp3Q it is 2.4 pixels wide, but IMHO that is even more likely to be diffraction limited, since there is no matrix-related alias blur with Foveon, so you would think the effect might start earlier. I have never seen an analysis on this point, though.

Shooting with f/4 and a good lens makes a lot of sense for ultimate detail comparisons, IMHO.

cheers

Thanks - I intended to shoot at f/5.6 but nailed the exposure early on with f/8 and got lazy...

-- hide signature --
 saltydogstudios's gear list:saltydogstudios's gear list
Ricoh GR Digital Ricoh GR Digital III Sigma DP2 Merrill Sigma DP3 Merrill Sigma dp3 Quattro +6 more
saltydogstudios Senior Member • Posts: 1,038
Re: Monochrome fine art on Sigma Cameras versus high res Bayer conversion?

OK so I mentioned in the previous post that I think Fuji's sharpening algorithms let it down when it comes to sharpness - so here's the Fuji run through Raw Therapee.

As I expected - this is considerably better.

I did enable Unsharp Mask and had to enable Chroma noise reduction (there's still considerable chroma noise). I didn't enable "Microcontrast".

X-Pro2 through Raw Therapee

As my previous experience (read blog here) - Raw Therapee brings the X-Trans MUCH closer to what a Foveon can do. Fuji's algorithms really do seem to work against them - it produces an interesting rendering, but for actual sharpness / detail - Raw Therapee is your best bet.

There is still considerable chroma noise that I don't know how to get rid of that seems to run in "lines" across the the image - horizontal and diagonal. I suspect this is a combination of things.

  • The Demosaicing algorithm is good in Raw Therapee - but not perfect.
  • The Denoising algorithms are tuned to Bayer patterns (as are most sharpening algorithms - just google "fuji wormies" for bad examples galore).
  • The fabric itself looks like "noise" to a denoising algorithm.

I'm sure some tweaking the sliders could tame it.

I chose Raw Therapee because it's said to have the best Fuji Demosaicing algorithms & Adobe's algorithms are said to mimic Fuji's so wouldn't be much different than SOOC.

Overall though - this is much better - Much more Foveon-like in its detail - even the red bit of lint is there!

Top: Sigma Quattro SOOC, Bottom: Fuji via Raw Therapee

The edge is still to the Quattro here, but the Fuji has a lot of promise.

If you convert the Fuji to monochrome so you don't have to worry about the chroma noise, throw some more sharpening ("microcontrast" setting in Raw Therapee), etc. you can get something very "sharp" - if perhaps overly "sharpened."

On the Quattro side, tweaking it in SPP, convert it to Monochrome mode (which tends to produce sharper output than color mode), export it to "Double" size and I suspect we'd be seeing some monster detail.

-- hide signature --
 saltydogstudios's gear list:saltydogstudios's gear list
Ricoh GR Digital Ricoh GR Digital III Sigma DP2 Merrill Sigma DP3 Merrill Sigma dp3 Quattro +6 more
saltydogstudios Senior Member • Posts: 1,038
Re: Monochrome fine art on Sigma Cameras versus high res Bayer conversion?

OK final post on this (especially since this isn't the Fuji forum) - but people wonder why the Fuji files get "wormies" when they're sharpened.

Look at how the (inset) Fuji X-Trans rendering turns the neatly ordered rows of fabric into an impressionistic painting, finding random shapes where there aren't shapes.

This is 100% down to Fuji's algorithms.

The Quattro "sees" the individual loops of fabric.

The Fuji - as proven above - can "see" them too. But their algorithm goes looking for random shapes instead.

-- hide signature --
 saltydogstudios's gear list:saltydogstudios's gear list
Ricoh GR Digital Ricoh GR Digital III Sigma DP2 Merrill Sigma DP3 Merrill Sigma dp3 Quattro +6 more
danski0224 Contributing Member • Posts: 971
Re: Monochrome fine art on Sigma Cameras versus high res Bayer conversion?

jande9 wrote:

How do you de-Bayer a camera?

Jan

Technically, one is removing the CFA (color filter array). The underlying sensor is monochrome.

The traditional Bayer sensor layout of RGGB pixels is still there without the CFA. If the RAW file from the converted camera is sent through standard processing software, the software doesn't know that the RGGB conversion process isn't needed.

One would need to use the software mentioned elsewhere in this thread to get the full benefit of a monochrome conversion.

The Sigma Quattro sensor has a 19.2 mp top layer, Merrill is 15 mp.

A 24mp Bayer based sensor has the pixels divided across RGGB, or approximately 6 mp each. The 50 mp Canon 5Ds/r works out to approximately 12.5 mp for each RGGB portion.

This is why the Quattro and Merrill sensors excel for monochrome use. Lots of detail for very little cost compared to getting a Bayer based camera converted and then using other processing software to get the full mp count from the sensor (or buying those other dedicated monochrome cameras from Leica or Phase One). Even the Sigma SD14 and 15 punch well above their weight- but in my opinion, the Quattro is much more usable.

-- hide signature --

Moments in Time, a work in progress.... https://www.flickr.com/gp/142423236@N08/965cs3

OP mikeodial Senior Member • Posts: 2,176
Re: Monochrome fine art on Sigma Cameras versus high res Bayer conversion?

danski0224 wrote:

jande9 wrote:

How do you de-Bayer a camera?

Jan

Technically, one is removing the CFA (color filter array). The underlying sensor is monochrome.

The traditional Bayer sensor layout of RGGB pixels is still there without the CFA. If the RAW file from the converted camera is sent through standard processing software, the software doesn't know that the RGGB conversion process isn't needed.

One would need to use the software mentioned elsewhere in this thread to get the full benefit of a monochrome conversion.

The Sigma Quattro sensor has a 19.2 mp top layer, Merrill is 15 mp.

A 24mp Bayer based sensor has the pixels divided across RGGB, or approximately 6 mp each. The 50 mp Canon 5Ds/r works out to approximately 12.5 mp for each RGGB portion.

This is why the Quattro and Merrill sensors excel for monochrome use. Lots of detail for very little cost compared to getting a Bayer based camera converted and then using other processing software to get the full mp count from the sensor (or buying those other dedicated monochrome cameras from Leica or Phase One). Even the Sigma SD14 and 15 punch well above their weight- but in my opinion, the Quattro is much more usable.

Your logic is the reason I will likely wait for the new Sigma FF which should be a new standard in BW digital tools .(Affordable ones at least)

 mikeodial's gear list:mikeodial's gear list
Leica X Vario Sigma SD1 Nikon D810 Sony a7R II Sigma sd Quattro H +11 more
Shawn67 Senior Member • Posts: 2,156
Re: Monochrome fine art on Sigma Cameras versus high res Bayer conversion?

saltydogstudios wrote:

OK so I mentioned in the previous post that I think Fuji's sharpening algorithms let it down when it comes to sharpness - so here's the Fuji run through Raw Therapee.

As I expected - this is considerably better.

I did enable Unsharp Mask and had to enable Chroma noise reduction (there's still considerable chroma noise). I didn't enable "Microcontrast".

X-Pro2 through Raw Therapee

As my previous experience (read blog here) - Raw Therapee brings the X-Trans MUCH closer to what a Foveon can do. Fuji's algorithms really do seem to work against them - it produces an interesting rendering, but for actual sharpness / detail - Raw Therapee is your best bet.

There is still considerable chroma noise that I don't know how to get rid of that seems to run in "lines" across the the image - horizontal and diagonal. I suspect this is a combination of things.

  • The Demosaicing algorithm is good in Raw Therapee - but not perfect.
  • The Denoising algorithms are tuned to Bayer patterns (as are most sharpening algorithms - just google "fuji wormies" for bad examples galore).
  • The fabric itself looks like "noise" to a denoising algorithm.

I'm sure some tweaking the sliders could tame it.

I chose Raw Therapee because it's said to have the best Fuji Demosaicing algorithms & Adobe's algorithms are said to mimic Fuji's so wouldn't be much different than SOOC.

Overall though - this is much better - Much more Foveon-like in its detail - even the red bit of lint is there!

Top: Sigma Quattro SOOC, Bottom: Fuji via Raw Therapee

The edge is still to the Quattro here, but the Fuji has a lot of promise.

If you convert the Fuji to monochrome so you don't have to worry about the chroma noise, throw some more sharpening ("microcontrast" setting in Raw Therapee), etc. you can get something very "sharp" - if perhaps overly "sharpened."

First, I also appreciate your discussion in this. I also really appreciate your posting the RAW files and then going back and try other software with the Fuji. If you are going for maximum sharpness that is really important with the Fuji. Your Rawtherapee image is better but Iridient Developer brings out the Fuji considerably more IMO.

Red lint is very obvious, and the pattern of the fabric is too without the noise in the RT version. If you have a Mac you should try Iridient Developer. It will also work for the Merrill files and is *much* quicker than SPP. If you are on Windows try Iridient X-Transformer which will covert the RAF files to DNGs that have already been demosaiced using Iridients algorithms, NR and sharpening.

Shawn

xpatUSA
xpatUSA Forum Pro • Posts: 15,531
Re: Monochrome fine art on Sigma Cameras versus high res Bayer conversion?

saltydogstudios wrote:

OK so I mentioned in the previous post that I think Fuji's sharpening algorithms let it down when it comes to sharpness - so here's the Fuji run through Raw Therapee.

As I expected - this is considerably better.

I did enable Unsharp Mask and had to enable Chroma noise reduction (there's still considerable chroma noise). I didn't enable "Microcontrast".

X-Pro2 through Raw Therapee

As my previous experience (read blog here) - Raw Therapee brings the X-Trans MUCH closer to what a Foveon can do. Fuji's algorithms really do seem to work against them - it produces an interesting rendering, but for actual sharpness / detail - Raw Therapee is your best bet.

There is still considerable chroma noise that I don't know how to get rid of that seems to run in "lines" across the the image - horizontal and diagonal. I suspect this is a combination of things.

  • The Demosaicing algorithm is good in Raw Therapee - but not perfect.
  • The Denoising algorithms are tuned to Bayer patterns (as are most sharpening algorithms - just google "fuji wormies" for bad examples galore).
  • The fabric itself looks like "noise" to a denoising algorithm.

I'm sure some tweaking the sliders could tame it.

I chose Raw Therapee because it's said to have the best Fuji Demosaicing algorithms & Adobe's algorithms are said to mimic Fuji's so wouldn't be much different than SOOC.

Overall though - this is much better - Much more Foveon-like in its detail - even the red bit of lint is there!

Top: Sigma Quattro SOOC, Bottom: Fuji via Raw Therapee

The edge is still to the Quattro here, but the Fuji has a lot of promise.

If you convert the Fuji to monochrome so you don't have to worry about the chroma noise, throw some more sharpening ("microcontrast" setting in Raw Therapee), etc. you can get something very "sharp" - if perhaps overly "sharpened."

On the Quattro side, tweaking it in SPP, convert it to Monochrome mode (which tends to produce sharper output than color mode), export it to "Double" size and I suspect we'd be seeing some monster detail.

Not saying you should but, in RawTherapee, I get good acutance by:

Open Contrast By Detail Levels and apply one 'click' of contrast on the '+' button, sometimes two.

Open Sharpening, select RL De-Convolution and tone it down quite a lot - maybe 0.57 radius, 63 amount, 15 damping, 40 iterations.

Open Impulse Noise Reduction and adjust the threshold as deemed necessary.

-- hide signature --

Ted

 xpatUSA's gear list:xpatUSA's gear list
Sigma DP2 Sigma DP2 Merrill Panasonic Lumix DMC-LX7 Panasonic Lumix DMC-G1 Sigma SD15 +17 more
Joris1632 Senior Member • Posts: 2,458
What a vast amount of pleasure ...

...  is to be derived from digital processing and comparisons of sensors and software.  And cameras. And lenses. Sadly some of us are not  able to share in these delights, we have a deficiency, perhaps congenital, which sees  PP as a price to be paid for digital convenience.  For the rest I wish you continued pleasure, perhaps an indisputable "best" may be achieved in your lifetimes, ..........  or maybe not?  

-- hide signature --

Joris1632

Lin Evans
Lin Evans Forum Pro • Posts: 17,638
Re: What a vast amount of pleasure ...
1

Joris1632 wrote:

... is to be derived from digital processing and comparisons of sensors and software. And cameras. And lenses. Sadly some of us are not able to share in these delights, we have a deficiency, perhaps congenital, which sees PP as a price to be paid for digital convenience. For the rest I wish you continued pleasure, perhaps an indisputable "best" may be achieved in your lifetimes, .......... or maybe not?

Once when I was young, I met Ansel Adams in San Francisco at a break in one of his presentations. I had the pleasure of discussing a few things including "Moonrise - Hernandez, New Mexico" and post processing. He said he was still tweaking the negative a full seven years after he snapped the shutter. "Post processing is definitely not just something digital shooters do, it's something which has been done for years and years with film photography - just not so easily... LOL

Best regards,

Lin

Keyboard shortcuts:
FForum MMy threads