Home · Register · Join Upload & Sell

Moderated by: Fred Miranda
Username  

  New fredmiranda.com Mobile Site
  New Feature: SMS Notification alert
  New Feature: Buy & Sell Watchlist
  

FM Forums | Leica & Alternative Gear | Join Upload & Sell

1       2              4       5       end
  

Leica Triple Resolution marketing vs reality. Another look.

  
 
LBJ2
Offline
• • • • •
Upload & Sell: On
p.3 #1 · p.3 #1 · Leica Triple Resolution marketing vs reality. Another look.


Fred Miranda wrote:
Sure! Itís an interesting discussion. Let us know what you discover something .


Just because I am interested and short of firing up Python, and buying a license and learning how to use the Pixinsight platform (Ha Ha) to generate the Delta values, that my own eyes can't see otherwise pushing +3 stops. I'm a bit at a loss. But have reached out to a few that might be able to tackle one way or the other.



Apr 06, 2024 at 02:16 PM
zhangyue
Offline
• • • • •
Upload & Sell: On
p.3 #2 · p.3 #2 · Leica Triple Resolution marketing vs reality. Another look.


OP, first of all. To avoid confusion, are you interested to look small raw DR advantage at the same target output size compare to large raw? That wonít happen. It will be about the same at best. Or you are interested to see at advantage of pixel level 100% view of small raw compare to large raw.

I remembered I did a test two years ago when I bought M11. You can try to do this: under exposure the same image by two stop, small raw and large raw. Disable all noise reduction, pull back exposure by 2-4 stop in Lightroom and Zoom in the black region 100%, you can see less noise in the same region with small raw.
Iíd assume people like Bill Cliff can analysis the raw file to determine the DR of small raw. If it is the same or very close to large raw that means small raw will have advantage of DR at pixel level.




LBJ2 wrote:
Just because I am interested and short of firing up Python, and buying a license and learning how to use the Pixinsight platform (Ha Ha) to generate the Delta values, that my own eyes can't see otherwise pushing +3 stops. I'm a bit at a loss. But have reached out to a few that might be able to tackle one way or the other.




Apr 06, 2024 at 07:02 PM
thrice
Offline
• • • •
Upload & Sell: Off
p.3 #3 · p.3 #3 · Leica Triple Resolution marketing vs reality. Another look.


This has been well-covered in the imx455 threads on astrophotography forums. The sensor (like all CMOS) does not have genuine analogue binning like CCD chips do. CMOS sensors in general do not offer binning because each pixel has its own amplifier and the binning occurs in 'software' (on the digital signal which includes read noise). The benefits of CMOS binning are much less than CCD binning but do exist. I have done binning on my mono astro camera which has imx455 simply because my seeing is not sufficient to justify the full resolution output. The noise/DR profile is not appreciably different, and in astrophotography you are pushing the files all the way to the limit.


Apr 06, 2024 at 07:48 PM
zhangyue
Offline
• • • • •
Upload & Sell: On
p.3 #4 · p.3 #4 · Leica Triple Resolution marketing vs reality. Another look.


Let me explain this in another way.
For small raw, there are two way to do this, letís simplify the problem say larger raw original size is 60M. And small raw is 15M. One way of doing this is pick 1 out of 4 pixel to create a 15M file. This way at pixel level, the SNR is the same because signal vs noise per pixel is not changed. However, original file have 4 time of area to collect information, if down size original file to 15M. Signal is simple summing, noise will root square sum. SNR of large raw will be better than small raw with 1 stop advantage.
Now letís talk about 2nd way of generate small raw. If camera at hardware level to do pixel binning, at first order, 4 pixel will be combined, so signal X 4, noise will be mostly random read noise (white, thermal noise), again it will RSS. So at pixel level, small raw will have better SNR, the same 1 stop.
So back to original question, Leica didnít disclose what they did, however given the fact they claim there is SNR improvement, I bet it is 2nd way which in line with my crude test result.
As you can see, this is a very useful and interesting feature for many who donít want deal with monster file but appreciate dynamic range and image rendering of images. It is like you have a A7s and A7r within the same body so that you can pick when to use which depend on your situation and mood



Apr 06, 2024 at 10:30 PM
highdesertmesa
Offline
• • • •
Upload & Sell: On
p.3 #5 · p.3 #5 · Leica Triple Resolution marketing vs reality. Another look.


zhangyue wrote:
...you can see less noise in the same region with small raw...


There is also less detail at the pixel level compared to resizing the L-DNG in PS. If you sharpen the smaller DNGs to try and regain some of the lost detail, does the noise increase? I would think so.



Apr 06, 2024 at 11:19 PM
CVickery
Offline
• • • •
Upload & Sell: On
p.3 #6 · p.3 #6 · Leica Triple Resolution marketing vs reality. Another look.


zhangyue wrote:
Let me explain this in another way.
For small raw, there are two way to do this, letís simplify the problem say larger raw original size is 60M. And small raw is 15M. One way of doing this is pick 1 out of 4 pixel to create a 15M file. This way at pixel level, the SNR is the same because signal vs noise per pixel is not changed. However, original file have 4 time of area to collect information, if down size original file to 15M. Signal is simple summing, noise will root square sum. SNR of large raw
...Show more

An interesting discussion. Question, Leica is using the second way, it would seem to me that the small raw would need to be a whole number fraction of the original file size...1/2 if two pixels are binned, 1/3 if three etc. The only way that I see to avoid this is to use some of the binned pixels for the adjacent binned pixel as well. In your 4 pixel example 1 pixel could also be used for the adjacent binned pixel. This should result in 25% more pixels overall: 18.75mp. Does his make any sense or is this discussion completely over my head?



Apr 06, 2024 at 11:21 PM
zhangyue
Offline
• • • • •
Upload & Sell: On
p.3 #7 · p.3 #7 · Leica Triple Resolution marketing vs reality. Another look.


highdesertmesa wrote:
There is also less detail at the pixel level compared to resizing the L-DNG in PS. If you sharpen the smaller DNGs to try and regain some of the lost detail, does the noise increase? I would think so.


Possible but is it the nature of 18M vs 60M in term of detail? Nowadays, I guess we could even do high resolution mode in software to reconstruct image with AI without introduce noise.

My description above on implementation is really theoretic first order assumption. Without known how Leica really implement this, (it do show all DNG are 14 bits) I can't say for sure how much difference will be compare to downsized full raw to small raw. I didnt spent much time on this myself then some simple test at first few day I got the camera because I really need is a 18M raw, it has higher priority than keep original "THE SAME" DR for me.

Having seen discussion here, each people may have different priority hence different view on this feature. I do think Leica is not necessarily misleading here depend on how you view/use this. I believe this is not a straightforward implementation that both Canon and Nikon actually didn't make them right. Nikon's mraw and sraw are not implemented in hardware binning but software resampling and truncate the original file, drop to 12 bit (so DR dropped) but file size is relative large compare to final resolution so penalty and trade off make it no longer attractive. Leica's M and S raw size are about resolution ratio, a real saving.

It would be really interesting to see Photon to photo's DR measurement for s/m/L raw for M11 to reveal all the insights.



Apr 07, 2024 at 12:39 AM
zhangyue
Offline
• • • • •
Upload & Sell: On
p.3 #8 · p.3 #8 · Leica Triple Resolution marketing vs reality. Another look.


CVickery wrote:
An interesting discussion. Question, Leica is using the second way, it would seem to me that the small raw would need to be a whole number fraction of the original file size...1/2 if two pixels are binned, 1/3 if three etc. The only way that I see to avoid this is to use some of the binned pixels for the adjacent binned pixel as well. In your 4 pixel example 1 pixel could also be used for the adjacent binned pixel. This should result in 25% more pixels overall: 18.75mp. Does his make any sense or is this discussion completely
...Show more

It is above my pay grade but I think there can be many ways to do this than simple integer binning. reuse certain pixel (as you listed), or not doing binning at certain area...?? I hope Leica can clarify the implementation publicly.



Apr 07, 2024 at 12:45 AM
freaklikeme
Offline
• • • • • •
Upload & Sell: Off
p.3 #9 · p.3 #9 · Leica Triple Resolution marketing vs reality. Another look.


zhangyue wrote:
It is above my pay grade but I think there can be many ways to do this than simple integer binning. reuse certain pixel (as you listed), or not doing binning at certain area...?? I hope Leica can clarify the implementation publicly.


I wouldn't hold your breath for it. I'm fairly certain Leica would be happy if no one ever questioned what they do in the background.



Apr 07, 2024 at 12:54 AM
highdesertmesa
Offline
• • • •
Upload & Sell: On
p.3 #10 · p.3 #10 · Leica Triple Resolution marketing vs reality. Another look.




zhangyue wrote:
Possible but is it the nature of 18M vs 60M in term of detail? Nowadays, I guess we could even do high resolution mode in software to reconstruct image with AI without introduce noise.

My description above on implementation is really theoretic first order assumption. Without known how Leica really implement this, (it do show all DNG are 14 bits) I can't say for sure how much difference will be compare to downsized full raw to small raw. I didnt spent much time on this myself then some simple test at first few day I got the camera because I
...Show more

Iím not sure a P2P analysis would be fair since the S and M-DNGs seem to resolve less detail than the L-DNG does when reduced to the same pixel dimensions in PS. It feels like weíd be comparing one file without noise reduction to another with it, even if that noise reduction was applied by a novel/non-traditional method.



Apr 07, 2024 at 08:16 AM
 


Search in Used Dept. 

highdesertmesa
Offline
• • • •
Upload & Sell: On
p.3 #11 · p.3 #11 · Leica Triple Resolution marketing vs reality. Another look.


One thing I havenít seen a comparison of is S and M-DNGs compare to the L-DNG but from the M11M instead of the M11. If/how the results differ might be informative.


Apr 07, 2024 at 08:37 AM
LBJ2
Offline
• • • • •
Upload & Sell: On
p.3 #12 · p.3 #12 · Leica Triple Resolution marketing vs reality. Another look.


zhangyue wrote:
OP, first of all. To avoid confusion, are you interested to look small raw DR advantage at the same target output size compare to large raw? That wonít happen. It will be about the same at best. Or you are interested to see at advantage of pixel level 100% view of small raw compare to large raw.

I remembered I did a test two years ago when I bought M11. You can try to do this: under exposure the same image by two stop, small raw and large raw. Disable all noise reduction, pull back exposure by 2-4 stop in
...Show more

Thank you. My interest is in attempting to confirm Mathphotographer's results: comparing DR/Noise between the full size 60MP DNG specifically to in-camera pixel-binned M/S files. Why? Because Mathphotographer's results in the video posted with both image comparisons and values seem to confirm Leica marketing concerning Triple Resolution, at lest with the Q3 in-camera pixel binned images tested/presented.

Based upon your comment and recommended simple approach to test the same, I think you understand my interest. I performed your recommended test approach as you described with the M11 and am currently evaluating the results. Currently trying to equalize zoom view/percent to allow for the differences in the amount of data between 18 and 36 MPs files and the 60MP file in LRC. *The larger file of course has a lot more data, so it appears more detailed, but I am not after detail, rather comparing DR/Noise.



Apr 07, 2024 at 09:44 AM
RustyBug
Offline
• • • • • • •
Upload & Sell: On
p.3 #13 · p.3 #13 · Leica Triple Resolution marketing vs reality. Another look.


zhangyue wrote:

It is like you have a A7s and A7r within the same body so that you can pick when to use which depend on your situation and mood


This ^ is the aspect that seems most attractive ... the inherent versatility it brings "in camera". That's not to suggest the versatility can't be "bested" by taking the full file and post processing it.

But, I think that it could be very valuable for an event photographer. For instance, a wedding reception party would be low light, high volume and small printing requirements. OTOH, formals, behind the scenes and others that could be candidates for heirloom quality prints would want the larger file.

Still, same body of choice.


The focus of the conversation to this point has been about splitting difference in IQ of in camera vs. post methods. Makes me wonder (back to the versatility thing), if there is an inherent performance difference (i.e. buffer, battery drain, etc.) when shooting a 100 or 1,000 L's vs. 100 or 1,000 in camera binned files. Does the process of binning "in camera" incur more time / energy than just capturing the full file. And, if so (either way), is it negligible ... or, does it start to "add up" in the course of the event.

So far, it seems like the pixel peeping comps have been all under very controlled, low volume, low pace conditions. I can't help but wonder how the difference (or not) plays out when things are going at Mach 2 and the chips are on the line ... L vs. M vs. S in camera performance diff's. Also, does the use / process of binning generate more / less heat (i.e. noise source) when things are uptempo?

Would be curious to see a performance test of say 100 consecutive L's vs. 100 consecutive M's / S's. That, and / or a test of battery life # shots for consecutive L's vs. M's vs. S's.

Granted, that's a significant undertaking, but ...

I guess my point is that while we take the path to maximum IQ in post ... is there a performance based consideration for the margin of difference in IQ from in camera vs. post that makes the in camera binning a quid pro quo attraction for performance (not just file size / storage) aspects, within the margin of use case demands.

I'm guessing that nobody has quantifiable info on this yet, but ... hmmmm



Apr 07, 2024 at 11:33 AM
zhangyue
Offline
• • • • •
Upload & Sell: On
p.3 #14 · p.3 #14 · Leica Triple Resolution marketing vs reality. Another look.


The key part is you need view all images at 100% for this comparison.
I didnít follow that YouTube YET, but if he claim S or M raw have beater DR at the same target output size, then I doubt that. Not possible.

In theory P2Pís result down sample all sensor to 12M or so for dynamic range plot. That is considered this down sampling advantage in raw program. That is why I want to see M11x s, m, L raw all down sampling to 12M or so and how they fight in DR.

I donít expect Leica can take advantage of full hardware binning to behave like a native 18M sensor but even they did this with binning at software level, I donít care. I really cared is can they have a similar DR to down sized large raw with much smaller raw file at cost of detail.

LBJ2 wrote:
Thank you. My interest is in attempting to confirm Mathphotographer's results: comparing DR/Noise between the full size 60MP DNG specifically to in-camera pixel-binned M/S files. Why? Because Mathphotographer's results in the video posted with both image comparisons and values seem to confirm Leica marketing concerning Triple Resolution, at lest with the Q3 in-camera pixel binned images tested/presented.

Based upon your comment and recommended simple approach to test the same, I think you understand my interest. I performed your recommended test approach as you described with the M11 and am currently evaluating the results. Currently trying to equalize zoom view/percent to
...Show more


Edited on Apr 07, 2024 at 12:21 PM · View previous versions



Apr 07, 2024 at 12:14 PM
zhangyue
Offline
• • • • •
Upload & Sell: On
p.3 #15 · p.3 #15 · Leica Triple Resolution marketing vs reality. Another look.


I got your point. Resolution advantage of down sample is 2nd consideration for me but it is a trade off for anyone want to use this small raw feature.
Even in dynamic range department, if S or M raw can match large raw, id consider I am lucky or a success for Leicaís implementation.

highdesertmesa wrote:
Iím not sure a P2P analysis would be fair since the S and M-DNGs seem to resolve less detail than the L-DNG does when reduced to the same pixel dimensions in PS. It feels like weíd be comparing one file without noise reduction to another with it, even if that noise reduction was applied by a novel/non-traditional method.




Apr 07, 2024 at 12:20 PM
LBJ2
Offline
• • • • •
Upload & Sell: On
p.3 #16 · p.3 #16 · Leica Triple Resolution marketing vs reality. Another look.


zhangyue wrote:
The key part is you need view all images at 100% for this comparison.
I didnít follow that YouTube YET, but if he claim S or M raw have beater DR at the same target output size, then I doubt that. Not possible.

In theory P2Pís result down sample all sensor to 12M or so for dynamic range plot. That is considered this down sampling advantage in raw program. That is why I want to see M11x s, m, L raw all down sampling to 12M or so and how they fight in DR.

I donít expect Leica can take advantage
...Show more

Understandably, many will not take the time to sit through and follow the video. Much more fun to be out shooting ! Besides, eye- balling L/M/S all looks pretty good at normal viewing distances anyway w/ of course L-DNG being the high resolution/detail king among the three.

However, If you are inclined to do so, then it will become clear how he came to his conclusions to include, M-DNG in-camera binned = DR/Noise best performance ratio compared to full L-DNG and the in-camera binned S-DNG. S-DNG is runner up to the M-DNG in his measurements.

Rightly, wrongly or somewhere in-between 😉



Apr 07, 2024 at 12:55 PM
highdesertmesa
Offline
• • • •
Upload & Sell: On
p.3 #17 · p.3 #17 · Leica Triple Resolution marketing vs reality. Another look.


zhangyue wrote:
I got your point. Resolution advantage of down sample is 2nd consideration for me but it is a trade off for anyone want to use this small raw feature.
Even in dynamic range department, if S or M raw can match large raw, id consider I am lucky or a success for Leicaís implementation.



I see your perspective. I'm hesitant to use the smaller DNGs because I would keep thinking about how much better the results would be at the pixel level if I shot a camera that had a native lower resolution sensor. For me to use S or M-DNG, I would want at least parity with a native sensor regarding sharpness/resolving power since we know we can get close enough to that by downsampling the L-DNG afterwards.

Perhaps when seeking a less digital, more filmic look the S and M-DNGs might be more appealing. Surely the smaller DNGs would produce less harsh results at the pixel level. I could see using the smaller DNGs when using an APO lens for casual portraits for example.

The biggest argument for never using the smaller DNGs, however, is that fact that it greatly increases the odds I will forget to change the resolution back to L-DNG for critical shooting. How disappointing it would be to shoot a once-in-a-lifetime landscape scene at 18mp when I could have shot it at 60mp. It would be nice if there was an option to enable an info screen that showed the current resolution setting at power-on provided it didn't slow startup time. I know I can see the setting elsewhere, but given my memory, I would prefer a warning screen



Apr 07, 2024 at 02:18 PM
highdesertmesa
Offline
• • • •
Upload & Sell: On
p.3 #18 · p.3 #18 · Leica Triple Resolution marketing vs reality. Another look.


I have now compared S/M/L-DNG on the M11M versus resizing in Photoshop:

My expectation was to see improved results with the S and M-DNGs from the M11M versus those from the M11 since there is no CFA, but results are identical to M11 comparisons.

I further experimented with the resizing options in Photoshop to see if I could duplicate exactly the resolving power (apparent sharpness) of S-DNG by resizing the L-DNG with a different resampling mode in PS:

Ė All the reduction/sharpening-based resample options still result in more resolved detail from the reduced L-DNG compared to the S-DNG.
Ė However, if I use the enlargement/smoothing resample option "Bicubic Smoother (enlargement)", then the result from resizing the L-DNG to the same size as the S-DNG gives identical resolving power.

It may seem counterintuitive to use an enlargement-based resampling method to downsample an image, but the results seem to suggest whatever Leica is doing in-camera on both M11 and M11M matches this way of downsizing in PS. Based on this, my theory is there is nothing going on except resizing in-camera in such a way as to reduce noise. Either the processor is fast enough to do this on the fly or there is a dedicated chip in the camera for this, IDK.



Apr 07, 2024 at 03:36 PM
Fred Miranda
Offline
Admin
Upload & Sell: On
p.3 #19 · p.3 #19 · Leica Triple Resolution marketing vs reality. Another look.


highdesertmesa wrote:
I have now compared S/M/L-DNG on the M11M versus resizing in Photoshop:

My expectation was to see improved results with the S and M-DNGs from the M11M versus those from the M11 since there is no CFA, but results are identical to M11 comparisons.

I further experimented with the resizing options in Photoshop to see if I could duplicate exactly the resolving power (apparent sharpness) of S-DNG by resizing the L-DNG with a different resampling mode in PS:

Ė All the reduction/sharpening-based resample options still result in more resolved detail from the reduced L-DNG compared to the S-DNG.
Ė However, if I
...Show more

Here is a comment by @Peter Figen regarding bicubic smoother:

https://www.fredmiranda.com/forum/topic/1654369/0&year=2020#16418375

I've been using as my default interpolation method in Photoshop, especially when manipulating files using rotation, cropping.



Apr 07, 2024 at 03:57 PM
zhangyue
Offline
• • • • •
Upload & Sell: On
p.3 #20 · p.3 #20 · Leica Triple Resolution marketing vs reality. Another look.


I watched video while waiting for my sons competition. I couldnít spend all the time to catch what he was doing. I can see a few things make no sense to me.

First of all, he is doing pixel level test, so that means small or medium raw will have advantage.
The conclusions he draw totally ignore the downsize advantage of large raw.

Having said that, even the methodology he draw the conclusion is flawed.

The image used makes no sense to me. He really should be using uniform gray image so that signal are uniformed like the fist example he shown.

The right way to do this is put lens cap on at lowest iso, take an image with exposure time required to clip the sensor with 18% gray card with S M and L raw.

Assuming all raw file are 14 bit, the clip digital code will be 16000. And do raw digger for three black images to get standard deviation of noise level at black. These will give true SNR.

If Leica do this right, you will small raw have lowest black noise, large raw has the worst. Then after resize, they will match.

The problem of his methodology is he was trying get standard divination of random image with pixel data all over the place. That is totally miss the fundamental concept. And if you check SNR of those theee image anre around 30dB or 5 stop. What is that? Donít we talk about at least 12-13 stop dynamic range?

I am not an expert on digital image signal process but do have strong background on ADC, DAC and fundamental understanding of signal processing. So I know the way he did is not right so is his conclusion.


LBJ2 wrote:
Understandably, many will not take the time to sit through and follow the video. Much more fun to be out shooting ! Besides, eye- balling L/M/S all looks pretty good at normal viewing distances anyway w/ of course L-DNG being the high resolution/detail king among the three.

However, If you are inclined to do so, then it will become clear how he came to his conclusions to include, M-DNG in-camera binned = DR/Noise best performance ratio compared to full L-DNG and the in-camera binned S-DNG. S-DNG is runner up to the M-DNG in his measurements.

Rightly, wrongly or somewhere in-between
...Show more



Apr 07, 2024 at 04:45 PM
1       2              4       5       end






FM Forums | Leica & Alternative Gear | Join Upload & Sell

1       2              4       5       end
    
 

You are not logged in. Login or Register

Username       Or Reset password



This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.