Fun with MM DNG - 4 XingPing bridge

[jonoslack]

Hi David

There is a thread over on the Lula forums about the D800 vs D800e. The guys there showed conclusively that with proper attention spent deconvolution sharpening, the acutance advantage of the D800e is negligible. The D800 provides the cleaner files.

Yes . . . but others have come to different conclusions.
Personally I think it's angels dancing on the head of pins!

 

[jonoslack]

http://farm9.staticflickr.com/8008/7250579166_ab26d74bb7_b.jpg[/IMG]

Jono, while this probably really won´t garnish any interest from any pirates out there, I´ll delete as soon as you say the word. I´m just amazed, and wanted to share.

LOL - I wondered who'd find this - it's from a chemists shop in Cardiff in the 19th century . . .but in the current context perhaps it's rather apropos.

Don't take anything down!

all the best

 

[fotografz]

Hi

A valiant attempt at smoothing the image but it doesn't work that well. Forget about the bike jaggies, it's the jaggies in trees that bother me. Not only are there still very obvious "spiral candle" type artifacts on thin branches but there is that patchy messiness on complicated overlapping foliage which screams aliasing. It's exactly what my NEX 3 does too.

I also don't really see the point of starting out with an aliased image then attempting to disguise it with various forms of blur. Why not just start off with a nice clean, artifact free anti-aliased image to start with and sharpen appropriately to restore the MTF around Nyquist.

There is a thread over on the Lula forums about the D800 vs D800e. The guys there showed conclusively that with proper attention spent deconvolution sharpening, the acutance advantage of the D800e is negligible. The D800 provides the cleaner files. The reviews of course draw a different conclusion because they don't do deconvolution.

I applaud Leica's mono initiative but there nothing that screams "digital" to me more than aliasing artifacts - cleanly rendered or "disguised".

Haven't a clue what you are talking about. Wish I could see it, but I don't ... However, I am viewing the original at 100% on a print. Perhaps you could do a screen grab from the Original size I loaded to Smug Mug, and point out what you are talking about. I do prints not web pics, so that the main criteria, and there I don't see anything like what you are mentioning.

BTW, not sharpening a file isn't disguising anything IMO. Conversely, perhaps sharpening is disguising a mushy image?

-Marc

 

[mjm6]

Marc,

I noticed the ailising artifacts in the first image Jono put up and commented on it (in the beard). However, it took viewing at 100% on the screen to see, and that equates to about 400% higher than I would normally print (360 ppi on the print normally).

My point when I saw it was that it is proof that the camera has high resolving capabilities... (higher than the normal camera offerings out there anyway) and that it just will take some care to avoid problems.

As you say above, it is absolutely a case of all cameras doing this depending on the magnification, if you are able to send an image sharp enough through the lens to the sensor, and if they camera doesn't smear the image with an AA filter.

So clearly, the lenses Jono used still have headroom before the sensor is more capable than the lenses, which makes me wonder what exactly Leica is thinking with the new 50mm Summi. At this point, I don't think they make a camera that can accept that level of sharpness (other than tech pan...)

However, a 36MP monochrome sensor... who knows where that will shake out. I bet close to the limits of these lenses.

Until the sensors are more capable than the lenses (so the lenses effectively act as an AA filter on their own) these ailising issues will occur, especially when looking at 100%.

A camera that doesn't suffer from ailising is going to fall into one of two camps; it has a fairly poor lens on it that doesn't meet the performance level of the sensor, or the sensor has a pretty strong AA filter applied, which ultimately results in a similar destruction of sharpness to avoid the issue. It could be a combination of both, too.

Ailising will clearly be a greater problem if there is no pixel averaging going on, which it appears is not (but there is some pretty strong debate on this as well). However, if there were much, I don't think the ailising would be visible as we have seen.

Regardless, I think it's a good problem to have, and something that can be dealt with most of the time without too much difficulty.


---Michael

 

[wentbackward]

Regarding the jaggies, the level of anti-aliasing is your own choice of how to interpret the conversion from an analogue world to digital. The artefacts seen in the samples above were excessive due to sharpening settings. Oversharpening the input causes the leaves to look mushy and blown out, whereas in fact they are not. I think there are some acacia trees in there and indeed they are difficult to resolve, (low contrast, small details, very random, pushing out of green into yellow). In the print they do show a nice random pattern, not mush as seen in colour on a regular camera.

The simple solution to these jaggies: Don't sharpen, upsize the image to output dimensions, sharpen for a given viewing distance, then print/render

 

[wentbackward]

Regarding printing, I can see the results can be improved (it was my first attempt at this image/camera/pure B&W, I did no test prints):
1. I still oversharpened the source/input/RAW conversion, I suspect (as Jono has pointed out already I think) no input sharpening is close to optimal on this image
2. The second improvement would be not sharpening the output enough, as I [perhaps wrongly felt] was losing tonality in areas that were close white.
3. There is some noise in the shadows which was subsequently enlarged, I think I can do some shadow specific noise reduction before enlargement.
4. Adding a little grain at the output stage would give the image more bite. I didn't add any grain last time.
5. The mounting card was from an emergency 'weekend' supplier. Two words sum it up. Expensive crap.

Those macro shots were shot through the glass btw as I wanted to bring the print home so stuck it in a frame.

Despite the problems, the image passed the 'wife' test today with a Wow, clearly that is all Jono's creative talent, not any technical aspect of the printer! Anyhow any image that makes her wow, would make someone else fall off their perch, so that's good!

Personally at 100x70cm I expected the details to pop at close to the same level as a Medium Format image. The image surely has the detail to pop, but it didn't quite do it for me.

I will print again with the adjusted settings, remount on better card, then post macro shots out of the glass.

 

[wentbackward]

Lastly, Jono ... you have incredibly stable hands for 1/250th at 75mm it is astonishingly good!

 

[m_driscoll]

Informative, educational, fascinating & entertaining!

Cheers, Matt

Zenfolio | Matt Driscoll

 

[davemillier]

Deleted, see next post

 

[davemillier]

Marc

i find with my 50 year old eyes that I need to don reading glasses to see these things clearly ;-)

Enthusiasts of the AA-less approach like to go on about sharpness and micro-contrast, the trouble is that in any sampled system you need to reduce MTF at Nyquist to suppress under-sampling artifacts, not increase it. Detail beyond Nyquist isn't detail, it's distortion. And the problem with distortion is that while sometimes it can fool the eye into seeing detail, other times what you see is horrible artifacts.

The problem with aliasing in landscapes for me isn't so much the obvious aliasing (reviewers always look for colour moire, don't they) but the effect it has on the "feel" of the rendering of grass and foliage from the little mentioned luminance aliasing. The example below even softened by your treatment still has obvious artifacts but these are just the gross symptoms. What I see is a kind of localised patchiness in rendering, depending on the subject. Sometimes you get nice sharpness and detail, sometimes you get strange broken, confused tones that look a bit like noise, other times you get a fabricated texture eg hair detail that renders like a canvas weave in patches. It is all so variable and random. What I find with my NEX is that some subjects look clean and sharp while other (the details of a broadleaf forest in winter for example) look noisy and messy. It's difficult to describe in words but it's basically the effect of rendering small detail as a fractured tile pattern with sharp geometric edges, then overlaying it on other small detail rendered as a fractured tile pattern with sharp geometrical edges. When you shrink this down to typical print sizes, the result looks nasty, noisy, unnatural to me. Apparently many people perceive it merely as "crisp".

Anyway, back to the easy stuff, a crop from your file:

By the way, here's an example from DPreview tests of the NEX. Compare the smoothness of the Olympus rendering to that of the NEX 5 on the curves of the coin (You need to select the "original" link and click to expand to full size). Hideous isn't it?


http://www.dpreview.com/galleries/286305481/photos/1900584/lum

Haven't a clue what you are talking about. Wish I could see it, but I don't ... However, I am viewing the original at 100% on a print. Perhaps you could do a screen grab from the Original size I loaded to Smug Mug, and point out what you are talking about. I do prints not web pics, so that the main criteria, and there I don't see anything like what you are mentioning.

BTW, not sharpening a file isn't disguising anything IMO. Conversely, perhaps sharpening is disguising a mushy image?

-Marc

 

[fotografz]

Marc

i find with my 50 year old eyes that I need to don reading glasses to see these things clearly ;-)

Enthusiasts of the AA-less approach like to go on about sharpness and micro-contrast, the trouble is that in any sampled system you need to reduce MTF at Nyquist to suppress under-sampling artifacts, not increase it. Detail beyond Nyquist isn't detail, it's distortion. And the problem with distortion is that while sometimes it can fool the eye into seeing detail, other times what you see is horrible artifacts.

The problem with aliasing in landscapes for me isn't so much the obvious aliasing (reviewers always look for colour moire, don't they) but the effect it has on the "feel" of the rendering of grass and foliage from the little mentioned luminance aliasing. The example below even softened by your treatment still has obvious artifacts but these are just the gross symptoms. What I see is a kind of localised patchiness in rendering, depending on the subject. Sometimes you get nice sharpness and detail, sometimes you get strange broken, confused tones that look a bit like noise, other times you get a fabricated texture eg hair detail that renders like a canvas weave in patches. It is all so variable and random. What I find with my NEX is that some subjects look clean and sharp while other (the details of a broadleaf forest in winter for example) look noisy and messy. It's difficult to describe in words but it's basically the effect of rendering small detail as a fractured tile pattern with sharp geometric edges, then overlaying it on other small detail rendered as a fractured tile pattern with sharp geometrical edges. When you shrink this down to typical print sizes, the result looks nasty, noisy, unnatural to me. Apparently many people perceive it merely as "crisp".

Anyway, back to the easy stuff, a crop from your file:

By the way, here's an example from DPreview tests of the NEX. Compare the smoothness of the Olympus rendering to that of the NEX 5 on the curves of the coin (You need to select the "original" link and click to expand to full size). Hideous isn't it?


lum: DMillier: Galleries: Digital Photography Review

Thanks.

IMO, that is taking the art of pixel peeping to new heights ... the crop you resorted to that shows such artifacts represents such a massive enlargement, inspected at nose-on-print distances with magnifying readers, as to enter a realm requiring new terminology ... inspection at the atomic level. Any further, and we'll be needing Electron Microscopes.

Sorry, but I do not see the practical effect on the print, including the areas you circled.

That said, while I like this camera, I still do not think it equals the impression of detail one gets using a M6 and film with the same lens directly printed on silver print paper ... that has been the elusive aspect of digital from the start ... although, few analog images could with-stand these pixel-peeping exercises either.

I also personally do not think these exercises line up very well with the intent of this camera.

-Marc

 

[jonoslack]

HI There
You're doing splendid work here - I think we are all being enlightened both as to the nature of the file, and also the way to print it.

Despite the problems, the image passed the 'wife' test today with a Wow, clearly that is all Jono's creative talent, not any technical aspect of the printer! Anyhow any image that makes her wow, would make someone else fall off their perch, so that's good!

Personally at 100x70cm I expected the details to pop at close to the same level as a Medium Format image. The image surely has the detail to pop, but it didn't quite do it for me.

I will print again with the adjusted settings, remount on better card, then post macro shots out of the glass.

Lastly, Jono ... you have incredibly stable hands for 1/250th at 75mm it is astonishingly good!

Thank you - My father always had really shaky hands - I think I'm very lucky, of course, it's partly technique, but coffee and alcohol don't seem to have much of an effect either. Long May It Last!

all the best

 

[jonoslack]

Anyway, back to the easy stuff, a crop from your file:

Nope, Sorry David - I can't get worked up about this - even at 100% on my big screen with reading glasses and 2 ft away. If you say it has a 'feel' on even small prints, I can't contradict you but then of course we are in the realms of subjective, rather than objective observation, and would need to standardise on printer / paper / software etc. etc.

What I think this thread has clearly demonstrated is that the default sharpening and base level clarity in Lightroom is too much for these files - Marc and wentbackward seem to have done a grand job here. It also suggests to me that the files really haven't had any processing done on them in LR.

All the best

 

[Vivek]

I think David owes a thanks for bring this up to evoke the responses from wentbackward and fotografz.

The file looks amazing!

There are possibilities that a polarizer would help to mitigate the extra stuff on the bicycle and the foliage and such but that is something to do with light and photography in general (film, digital, whatever) than the MM.

 

[davemillier]

Marc, Jono

Just to be clear, are you saying that (on screen at least, I haven't printed this image), that the areas I have circled don't jump out at you instantly as horribly jagged zig zag lines?

To my eyes, those branches, vines or whatever they are, that should render as smooth tubes, look like twisted fibres or rope because of the aliasing.

This is the problem I see with Foveon based cameras which have no AA filter. Any fine lines at the right (wrong?) angle reproduce as a jagged line or look like twisted rope. As you (should!) be able to see here...

(I appreciate that the Leica files have more resolution than the Sigmas, so the question of scale comes into it).

Did you look at the NEX 5 files from DPreview I linked to as a clearer demonstration of the problem? What did you think of those?

Nope, Sorry David - I can't get worked up about this - even at 100% on my big screen with reading glasses and 2 ft away. If you say it has a 'feel' on even small prints, I can't contradict you but then of course we are in the realms of subjective, rather than objective observation, and would need to standardise on printer / paper / software etc. etc.

What I think this thread has clearly demonstrated is that the default sharpening and base level clarity in Lightroom is too much for these files - Marc and wentbackward seem to have done a grand job here. It also suggests to me that the files really haven't had any processing done on them in LR.

All the best

 

[jonoslack]

I think David owes a thanks for bring this up to evoke the responses from wentbackward and fotografz.

hi Vivek
Thanks for pointing that out.

Absolutely. I quite agree. I think it's generated a really useful discussion and given us all a better understanding of the files. Worth mentioning that David and I go way back to the early days of dPreview, and he has always been meticulous (and a good guy)

Thank you David

I think that it IS a matter of scale, and that at this level you have a straigthforward choice. Either some jagged or some blurred detail via an AA filter (whether that be hard or software based). I can't see an alternative

However, backing off the clarity, and zeroing the input sharpening certainly reduces the problem significantly.

All the best

 

[wentbackward]

...However, backing off the clarity, and zeroing the input sharpening certainly reduces the problem significantly...

Yes, good point Jono, the clarity slider will also introduce jaggies.

It seems there are a few people interested in this stuff so I'll blab on a bit if that's ok. Positive clarity values can be emulated in photoshop using an unsharp mask with settings of between 30-75% and adjusting the radius slider up and down. So in effect it is also a form of sharpening, or generalised as local contrast adjustments. Understanding this as local contrast (i.e. the contrast between a few pixels), you can begin to imagine how one might cause aliasing to occur. Indeed these techniques are also used in the darkroom or built right into film chemistry.

Sharpness is a bit of a tech-fest and it's easy to get bogged down in it. The mind/eye system is amazing and one picture with a few sharp details allows humans to perceive the whole image as sharp. It's why I love watercolours so much and also a good reason why the leaves don't matter so much in this image (for most viewers). Of course some will want Gursky sharp details with your nose pressed against the glass. I have to admit it's wonderful exploring large prints and seeing all the small details. It's my dream to see a real large Gursky.

Paul

 

[jonoslack]

HI Paul
I guess I'd realised that Clarity was another form of sharpening - I think one should pass an exam in restraint before using it (especially with the MM).

Thank you so much for your contribution to the thread - I think most of us have learned something (I know I have).

all the best

 

[davemillier]

According to a blog article I was reading the other day, LR4 clarity slider is much changed from previous versions. It is supposed to be "adaptive" now and will permit much higher values without the distortions of earlier LR versions.

There is also a trick that LR does when sharpening, you might find useful. If you hold down the alt key while hovering over the mask slider, you get a visual indication of exactly where sharpening is being applied and not applied throughout the image. Here's the explanation: Lightroom Tips - Visual Sharpening Mask

Yes, good point Jono, the clarity slider will also introduce jaggies.

It seems there are a few people interested in this stuff so I'll blab on a bit if that's ok. Positive clarity values can be emulated in photoshop using an unsharp mask with settings of between 30-75% and adjusting the radius slider up and down. So in effect it is also a form of sharpening, or generalised as local contrast adjustments. Understanding this as local contrast (i.e. the contrast between a few pixels), you can begin to imagine how one might cause aliasing to occur. Indeed these techniques are also used in the darkroom or built right into film chemistry.

Sharpness is a bit of a tech-fest and it's easy to get bogged down in it. The mind/eye system is amazing and one picture with a few sharp details allows humans to perceive the whole image as sharp. It's why I love watercolours so much and also a good reason why the leaves don't matter so much in this image (for most viewers). Of course some will want Gursky sharp details with your nose pressed against the glass. I have to admit it's wonderful exploring large prints and seeing all the small details. It's my dream to see a real large Gursky.

Paul

 

[wentbackward]

Hi Dave, Useful tip on the sharpening!

I can concur, the clarity slider is definitely more complex than I described and many people hate the new Algo. LR for me just seems to do everything wrong. Clearly this is a familiarity issue on my part as are problems with the new algo for many people (I can't get to grips with the old or new one). At least the manual approach in PS is what it is, a bit crummy, but reliable (I used to be able to do a USM using layers but forget how now, it took me a long time to trust the USM wouldn't be changed suddenly) Old habits do die hard but hopefully a little nuts'n'bolts understanding is helpful.

There are some cool edge detection and filtering algo's out there, I'm sure things will improve immensely for large format printers, over the coming years. Sandy McGuffog always has useful knowledge on his blog about how digital works.