Seeing any Posterization issues with A7r ll?

[Guy Mancuso]

Illiah, Tim and others appreciate you guys being here to help us out through this one. This one we needed the science.

 

[Amin]

The file rendered into ProPhoto RGB is acceptably clean, needs only minor retouch on small spots, done in less than 4 minutes. The conversion to Adobe RGB, however, is the whole different story. Concentrating on the camera and shooting conditions shifts the accent into the wrong direction here.

This is good information, thanks Iliah.

Some have reduced this to an issue of whether the RAW data is posterized or not. It is posterized (according to those who have seen it), but any camera can have posterized data under the right conditions, so I think it's not right to boil the entire issue down to that without taking into account the other factors.

 

[timparkin]

The file rendered into ProPhoto RGB is acceptably clean, needs only minor retouch on small spots, done in less than 4 minutes. The conversion to Adobe RGB, however, is the whole different story. Concentrating on the camera and shooting conditions shifts the accent into the wrong direction here.

Are you talking about a few (5-10) very small edges here and there (in the order to ~10 pixels) where there is visible posterisation? How small are we talking and are you just softening those edges to fix things?

 

[timparkin]

Illah, Tim and others appreciate you guys being here to help us out through this one. This one we needed the science.

Just processed the D810 and A7R files in Lightroom with 60,0.7,70 sharpening. The Sony is way worse than the Nikon in terms of what I think is being termed orange peel. However, when I process them as raw tiffs in RPP and then apply sharpening in lightroom (the same settings) and I see very little difference.

The main difference I can see is that the D810 has evidence of an antialias filter whereas there is evidence antialiasing on the A7R image.

Hence my only conclusion is that Lightroom/ACR demosaics or sharpens raw files differently based on manufacturer.

Tim

 

[ErikKaffehr]

Hi,

We need to keep things apart. There are naturally gaps in the data as Sony uses 11-bit worth of data (1:2048) distributed over 14 bits range (1:16384) at low levels the gaps are small with increasing levels the gaps are larger. The idea is the natural variation of light called "shot noise" will mask the gaps. Another way to see it that the precision of data is low so the natural spread should be larger than the gaps.

Now, there are some odd gaps in the data. This occurs when data are shifted (multiplied by a fractional number). What could cause this? Lens corrections may be a cause. Not very likely that it could cause problems.

Best regards
Erik

David, I don't agree that the boundaries of the issue have been defined.

We don't know to what extent (if any) Lloyd's use of lens compensation settings affected the gaps in the raw data.

We don't know to what extent (if any) another camera with uncompressed or losslessly compressed 14-bit raw would have shown similar effects with the same subject, degree of relative underexposure, and color management.

 

[ErikKaffehr]

Hi "Pegelli",

I have also read the Bruce Frazer/Jeff Schewe books and also some other stuff. I am also a bit familiar with Ed Grangers work on SQF.

I have been trough many cameras, but many of my best images were taken with a Sony Alpha 700. That camera had merely 12 MP but those 12 MP were good enough for very good A2 prints.

Now, skipping forwards, the Sony Alpha 900 arrived with 24 MP on full frame. I did compare the files from the Alpha 700 and the Alpha 900, and the Alpha 900 certainly played in another division. But, comparing A2 prints there was not much of a difference.

From the A900 I progressed to a Phase One P45+. Again, I found that the P45+ files were much sharper, but I never really have seen a difference in A2-size prints. Now, A2 matters to me. That is my standard print size.

I frequently use a program called Imatest. It can calculate SQF for different print sizes. I applied Imatest to my images from the P45+ and the Sony Alpha and also thrown in some images from Imaging resource and DPReview. What I have found that SQF with Imatest's standard sharpening was essentially identical for all images.

What I have realised is that we should sharpen so medium frequency detail is near 100% MTF. Now looking at actual pixels with focus on extinction resolution and related MTF. So we push sharpening on small detail that does not really affect our vision and ignore sharpening that actually makes prints better.

Bruce Frazer and Jeff Schewe have been there and done that. Jeff sharpening for faces and landscape in Lightroom may be a good starting point, that may give better results than achieving maximum MTF on actual pixels.

This article may be worth a look: http://echophoto.dnsalias.net/ekr/i...y-i-cannot-see-a-difference-in-a2-size-prints

BTW, I got my A7rII today.

Best regards
Erik

Hi Erik, I think no sins are committed so no stones need to be thrown. But you raise an interesting point. I've learned most of my sharpening from reading a lot of material by Bruce Frazer and Jeff Schewe who tell us to do as little "capture sharpening" as possible and then do your creative and final sharpening specifically for the output size and medium you're preparing the file for (print or screen). Since the sharpening settings in Lightroom (and I think C1 as well) are "capture sharpening" I do those very mildly and not far from the default settings. Just enough to judge if a file is sharp enough to further process or not. This way I hardly ever have problems with exaggerating noise and/or orange peel effect. After export from LR to prepare the file for printing or web showing I do a final round of noise reduction (only when required) followed by output sharpening specific for the output size and medium. I like the results this workflow gives me and wonder what advantages you (and may be the others you mention) see in doing a very aggressive capture sharpening, since that tends to exaggerate noise and other artifacts that are present in the file. What do you think I am missing by doing only a very mild capture sharpening?

 

[Guy Mancuso]

Congrats on the A7rII. I think you will really like it. Do me a favor forget all this stuff and just go out and shoot the dang thing and have fun. End of day that is what counts. And you better post some images soon. LOL

 

[pegelli]

This article may be worth a look: http://echophoto.dnsalias.net/ekr/i...y-i-cannot-see-a-difference-in-a2-size-prints

BTW, I got my A7rII today.

Best regards
Erik

Thanks for the explanation Erik, and an interesting article. My "easy" conclusion is that I'm not losing a lot of SQL with less aggressive capture sharpening for my camera's (for the A7 and A7r it's 1 point while about 5 are needed for a visible difference). I also assume most of these tests were done at low iso and low noise, so for higher iso pictures with more noise the orange peel effect when aggressively capture sharpening might actually lower the SQF (pure assumption from my side, has this been tested?) vs. more moderate capture sharpening. The counter argument might be that the orange peel actually increases our perception of sharpness (like in the old film days film grain could also increase the perception of sharpness).

But let me agree with Guy, it might all be a bit academic, especially now that you got your A7Rii and are probably dying to go out and test it. For me tomorrow is also a shooting day (day 240 on Dyxum, so I've got an old Porst 240/4.5 attached to my A850 ready to go) and I'm looking forward to some of your results with the A7Rii. I was allowed to borrow one for a couple of hours last Sunday and liked it a lot, hope it will be the same for you.

 

[Iliah Borg]

Are you talking about a few (5-10) very small edges here and there (in the order to ~10 pixels) where there is visible posterisation? How small are we talking and are you just softening those edges to fix things?

Some were a little more pronounced, like about 50 red pixels long (100 pixels, if after demosaicking)

 

[ErikKaffehr]

Hi Iliah,

I guess you posted the wrong image?

Best regards
Erik

Some were a little more pronounced, like about 50 red pixels long (100 pixels, if after demosaicking)

 

[Iliah Borg]

Hi Iliah,

I guess you posted the wrong image?

Best regards
Erik

Dear Erik,

It is a selection (over the coordinates shown in the title bar) that has abnormal pixel levels distribution, indicating posterization.