Pixel binning

[adli]

Advertisement (gone after registration)

 

Leica have stated that all 60MP are used for all three resolutions. The reduction is performed in fw/hw post to capturing the picture.

This means that as indicated by several here, there is no real point of shooting at 39 or 18 MP from and image quality point of view. Shoot 60 MP and reduce in post later if needed will yield same or better result if properly performed.

 

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[kirkmc]

On 2/2/2022 at 8:31 PM, BJohn said:

I mean they most certainly do some sort of magic in the firmware. Normally you cannot (at least to my knowledge) easily reduce a raw image with pixel binning (like 4:1). So what I also wonder, whether we have some sort of advantage to use the smaller DNGs or if it actually makes no difference to during post-processing (like picking the 60MP DNG and resize the image later on).

I think the point of this is to allow for more images in burst mode, so they larger files don't overload the buffer, as well as fitting more photos on an SD card (though that's secondary, since you can always change cards). Also, some people don't want the files to be that big, because of how they may transfer them. 

Interestingly, the raw files are smaller than the files for the 47 Mp Q2M; sample raw files I've seen for the M11 range from about 50-80 MB, whereas the Q2M files are around 85 MB. Q2 color files are around the same size; why would they be larger than 60 Mp files? 

[SrMi]

9 minutes ago, kirkmc said:

I think the point of this is to allow for more images in burst mode, so they larger files don't overload the buffer, as well as fitting more photos on an SD card (though that's secondary, since you can always change cards). Also, some people don't want the files to be that big, because of how they may transfer them. 

Interestingly, the raw files are smaller than the files for the 47 Mp Q2M; sample raw files I've seen for the M11 range from about 50-80 MB, whereas the Q2M files are around 85 MB. Q2 color files are around the same size; why would they be larger than 60 Mp files? 

Maybe because M11’s DNGs are lossless compressed, while Q2’s DNGs are uncompressed?

[kirkmc]

7 minutes ago, SrMi said:

Maybe because M11’s DNGs are lossless compressed, while Q2’s DNGs are uncompressed?

Ah, i didn’t realize that. Why don’t they offer that on the Q2M? Perhaps the processor can’t handle the compression?

[M11 for me]

vor 9 Stunden schrieb adli:

 

Leica have stated that all 60MP are used for all three resolutions. The reduction is performed in fw/hw post to capturing the picture.

This means that as indicated by several here, there is no real point of shooting at 39 or 18 MP from and image quality point of view. Shoot 60 MP and reduce in post later if needed will yield same or better result if properly performed.

 

I wonder now if its the correct understanding. You mention 39Mpix and things might be mixed up:

In L-DNG you shoot with full 60MPix. In M- resp. S-DNG you get 36 resp. 18MPix. So when do the 39MPix come into play? Besides L- or M- or S-DNG you can use 1.3x or 1.8x Crop. In that instance you always start from the highest resolution regardless the Crop you choose. This given when you shoot S-DNG at at the same time you choose 1.3x Crop then the camera immediately changes to the full 60MPix and on the LCD it shiws you the crop frame. That croped area is then 39MPix out of the 60MPix. When you import that now into Lightroom and you look at the sizes of your images you see akway the full 60MPix plus the croped area. You can keep it or reject it. So its the same as in the Q/Q2.

Just to conclude: When during shooting you go out of your crop-mode then the camera is back to your choosen L- or M- or S-DNG.

There is one thing that confuses me a bit: I just made a few tests in my studio: Even though the 1.8x Crop should be the same size as the L-DNG or the 1,3x Crop, the 1.8 is smaller. Why? Has maybe to do with the metering resulting in different amount of data. I. ust say that it is night here. I will have to tst in daylight tomorrow.

[FlashGordonPhotography]

I like doing things in camera. Was out today with the M11 and used 60 and 18 interchangeably throughout the day. One less step to worry about when I get home. Plus it works with the "why am I taking this photograph?" mentality I try to use when shooting. I've spent months of my life in darkrooms and Lightroom. I'd rather be shooting than processing so changing resolutions works for the way I like to shoot. Then again I also thend to have a fixed white balance rather than auto.

Gordon

[aristotle]

Advertisement (gone after registration)

Nobody has posted (to my knowledge) what Leica is actually doing with the reduced resolution modes, but in general terms, you do in fact get better noise performance if you combine sensor data prior to A/D conversion.  Adding two smaller sensor sites prior to sampling can approach the SNR of a single sensor site with twice the area.  Adding the outputs of two smaller sites after A/D conversion does not approach the SNR of the single larger site.   It is absolutely not the same thing as doing post-processed resolution reduction, no matter how sophisticated.  As to the physical geometry of what sensor sites get combined with which others, that's a detail that we'll not likely find out for the M11.  And, many (or most) might not notice the improvement with a "binned" output and a post-processed output at the same resolution for typical images....but there is a difference (unless it's implemented poorly...)

[SrMi]

35 minutes ago, aristotle said:

Nobody has posted (to my knowledge) what Leica is actually doing with the reduced resolution modes, but in general terms, you do in fact get better noise performance if you combine sensor data prior to A/D conversion.  Adding two smaller sensor sites prior to sampling can approach the SNR of a single sensor site with twice the area.  Adding the outputs of two smaller sites after A/D conversion does not approach the SNR of the single larger site.   It is absolutely not the same thing as doing post-processed resolution reduction, no matter how sophisticated.  As to the physical geometry of what sensor sites get combined with which others, that's a detail that we'll not likely find out for the M11.  And, many (or most) might not notice the improvement with a "binned" output and a post-processed output at the same resolution for typical images....but there is a difference (unless it's implemented poorly...)

Leica has said that going to a lower resolution means less noise in the shadows (better DR), but the same effect can be achieved with 60MP if you reduce the resolution in post. See talk at 43:10:

 

[scott kirkpatrick]

I don't know the details of the algorithm Leica chose and how they arrived at their choices of dimensions for the reduced DNG files that they produce.  Just as with super-resolution (multiple shots taken at half pixel spacing in cameras with IBIS), the basic trick is to create a series of transformations in which the data remains at the quality that it had in the DNG or DNGs that you start with.  So first, you translate each pixel signal into three 14-bit R, G, B components -- you know the transformation because the Bayer filter characteristics are part of every DNG file.  Then, without throwing any bits away, you combine them to create representative signals at the desired new lattice positions.(These averaged pixels actually have three 14 bit components, like a TIFF file.)  These are then transformed back as if there was a Bayer filter at each of the new "virtual" reduced lattice of sites.  The result is still 14 bits in depth at each site, and we have software to work with such files just as if they were original data.  Since Leica has probably used 16 bit words for all of the data in this calculation, and tries to underexpose a bit to protect highlights, there is the possibility of averaging to reduce noise in the shadows by a lot, and highlight errors by a little as well.  And maybe, being good engineers, they do a better job than Lightroom, which has to develop algorithms that work for everybody, not just for Leica.

Edited February 5, 2022 by scott kirkpatrick

[Robert Blanko]

So far I have not seen any comparison which would provide evidence for a significant improvement of image quality due to the in-camera downsampling over downsampling in post processing.

Several weeks after launch of the M11 it is thus a fair assumption, that there is no significant advantage (apart from saving memory or increased burst rate).

And this seems to be consistent with Leica‘s own statement in the video referred to in previous posts.

 

If I would use an M11, I would not see the point of using in-camera downsampling for my purposes. On the contrary, the imagination to lose details that later on could have been interesting would drive me crazy. There are these moments when you watch your photos at home on a large screen and see interesting details justifying a crop but that you did not notice when taking the photo…

[JPP1]

14 hours ago, M10 for me said:

There is one thing that confuses me a bit: I just made a few tests in my studio: Even though the 1.8x Crop should be the same size as the L-DNG or the 1,3x Crop, the 1.8 is smaller. Why? Has maybe to do with the metering resulting in different amount of data. I. ust say that it is night here. I will have to tst in daylight tomorrow.

Probably because the embedded previews are based on the cropped size.

[aristotle]

8 hours ago, SrMi said:

Leica has said that going to a lower resolution means less noise in the shadows (better DR), but the same effect can be achieved with 60MP if you reduce the resolution in post. See talk at 43:10:

 

Well, that pretty much settles that

[Thraph]

On 2/2/2022 at 8:52 PM, Photoworks said:

Just don't think about it.

I have done lots of testing and this is the best sensor on an M body.

Noise and DR in lower resolution is and advertising gimmick , At 60 MP you get excellent performance , at 36MP you get a smaller file.

Thanks it!

Hi, If you look at San Reid review, it appears that at higher ISO there is a difference btw 60 and 36 MP, also when compared to M10-R Resized. Obviously, those differences are tiny. I agree that 60mp is hugely qualitative, and I don't need IBIS to make sharp and in focus pictures for street photography (okay ... paying attention to be steady  ). Actually I don't see any difference in the 60mp and 36mp for the pictures I take, as Photoworks at Average-Low ISO. At higher ISO 3600+ I see the same as Sean Reid ... but only when I crop 1:1; looking at the whole picture, it is like the same (on screen)

[Herr Barnack]

On 2/2/2022 at 2:52 PM, Photoworks said:

Just don't think about it.

I have done lots of testing and this is the best sensor on an M body.

Noise and DR in lower resolution is and advertising gimmick , At 60 MP you get excellent performance , at 36MP you get a smaller file.

Thanks it!

Maybe I'm a simple minded kind of guy, but IMHO the above matters more than obsessive ruminating and hair splitting. 

Leica has labored long and hard to create a digital M masterpiece - let us be glad for that and enjoy the fruits of their labor.  🙂

Edited February 5, 2022 by Herr Barnack

[adli]

On 2/4/2022 at 9:23 PM, M10 for me said:

I wonder now if its the correct understanding. You mention 39Mpix and things might be mixed up:

In L-DNG you shoot with full 60MPix. In M- resp. S-DNG you get 36 resp. 18MPix. So when do the 39MPix come into play? Besides L- or M- or S-DNG you can use 1.3x or 1.8x Crop. In that instance you always start from the highest resolution regardless the Crop you choose. This given when you shoot S-DNG at at the same time you choose 1.3x Crop then the camera immediately changes to the full 60MPix and on the LCD it shiws you the crop frame. That croped area is then 39MPix out of the 60MPix. When you import that now into Lightroom and you look at the sizes of your images you see akway the full 60MPix plus the croped area. You can keep it or reject it. So its the same as in the Q/Q2.

Just to conclude: When during shooting you go out of your crop-mode then the camera is back to your choosen L- or M- or S-DNG.

There is one thing that confuses me a bit: I just made a few tests in my studio: Even though the 1.8x Crop should be the same size as the L-DNG or the 1,3x Crop, the 1.8 is smaller. Why? Has maybe to do with the metering resulting in different amount of data. I. ust say that it is night here. I will have to tst in daylight tomorrow.

 

You are using a typo to make a statement.....

[Photoworks]

On 2/5/2022 at 6:36 PM, Herr Barnack said:

Maybe I'm a simple minded kind of guy, but IMHO the above matters more than obsessive ruminating and hair splitting. 

Leica has labored long and hard to create a digital M masterpiece - let us be glad for that and enjoy the fruits of their labor.  🙂

I agree with you that all this advancements are great, we just need to understand the functionality and make and informed decisions on what to use.

for me the resolution of the M10-P was a little to low, upgraded to M10-R and I was happy. I will shoot the M11 mostly at 60MP.

Keep in mind that over time software improves and you can reprocess images from few years ago much better today. It will be the same in a few years with downsizing and noise handling in capture one or Lightroom.

[mujk]

Digital M cameras are unique among other modern cameras in the sense that it is possible, to a fairly high degree, to understand what the camera does for a certain combination if settings. This is because Ms are still mostly manual cameras. I think this is one of the M features that make it feel like a closer contact between the photographer and the subject, compared to having all the automation stuff in between.

Of course there are some things Leica will not tell how they do it in any detail, like in-camera lens corrections, probably because they think it is a business secret. But in this case I would find it useful if they told how pixel binning works in the M11, for example to understand the source of possible artefacts.

Edited February 11, 2022 by mujk

[IkarusJohn]

I remain confused about how this works in practice.

The sensor has 60MP, ie, 60 million sensel sites.  It also uses a Bayer colour filter array, which assigns 2 green, one red and one blue filter to each cluster of 4 sensels.  There is then a demosaicing, where each cluster is “interpolated” to result in a “colour” for each pixel.  The reason the 18MP Monochrom provided such fantastic resolution and clarity of its files is the Bayer CFA and the demosaicing was excluded - you simply got the intensity of light reading on each sensel site.

So, with that in mind, how do we go from 60MP to 36MP to 18MP.  Clearly this is not by dumping the reading from every second sensel site, or some other fraction.  The CFA is a physical filter over the sensor … or is it?

Surely, however this works, the starting proposition is that the sensor records the entire 60MP, and the interpolation/demosaicing required to make sense of the output after the Bayer CFA is applied to the output whichever “resolution” is selected.  @FlashGordonPhotography tells us (if I understood him correctly) that while there is a slight improvement in the 36MP files, effectively there is no loss in image quality for any of the three options, other than resolution.  I’m not sure he suggests that there is the gain one would expect from larger sensel sites, but he gets close to saying there is only marginal improvement in output from the Sony 50c sensor in the X1D II.

I guess I can see the benefit in handling smaller files, and having the option to crop for larger files, but otherwise I really struggle to see why this camera has a 60MP sensor …

PS - @scott kirkpatrick’s explanation at #29 above seems the most logical.  Does this mean that in future, we will have massive MP sensors (potential 100MP) where you set your preferred output file size, with no apparent loss in dynamic range, increase in noise or any other advantage or disadvantage in file size, other than more options in resolution?

Edited February 23, 2022 by IkarusJohn

[Advertisement]

Hallo,

Look here: Interesting Leica gear

[SrMi]

4 minutes ago, IkarusJohn said:

I remain confused about how this works in practice.

The sensor has 60MP, ie, 60 million sensel sites.  It also uses a Bayer colour filter array, which assigns 2 green, one red and one blue filter to each cluster of 4 sensels.  There is then a demosaicing, where each cluster is “interpolated” to result in a “colour” for each pixel.  The reason the 18MP Monochrom provided such fantastic resolution and clarity of its files is the Bayer CFA and the demosaicing was excluded - you simply got the intensity of light reading on each sensel site.

So, with that in mind, how do we go from 60MP to 36MP to 18MP.  Clearly this is not by dumping the reading from every second sensel site, or some other fraction.  The CFA is a physical filter over the sensor … or is it?

Surely, however this works, the starting proposition is that the sensor records the entire 60MP, and the interpolation/demosaicing required to make sense of the output after the Bayer CFA is applied to the output whichever “resolution” is selected.  @FlashGordonPhotography tells us (if I understood him correctly) that while there is a slight improvement in the 36MP files, effectively there is no loss in image quality for any of the three options, other than resolution.  I’m not sure he suggests that there is the gain one would expect from larger sensel sites, but he gets close to saying there is only marginal improvement in output from the Sony 50c sensor in the X1D II.

I guess I can see the benefit in handling smaller files, and having the option to crop for larger files, but otherwise I really struggle to see why this camera has a 60MP sensor …

You cannot expect any gain from larger sensel sites when comparing the same output size. However, if you compare at the pixel level, you will see less noise with larger pixels or M-DNG vs. L-DNG.
All 60MP-s are read out in every resolution mode and then processed in the camera for appropriate resolution.

[IkarusJohn]

1 hour ago, SrMi said:

You cannot expect any gain from larger sensel sites when comparing the same output size. However, if you compare at the pixel level, you will see less noise with larger pixels or M-DNG vs. L-DNG.
All 60MP-s are read out in every resolution mode and then processed in the camera for appropriate resolution.

What?  If there are larger sensel sites with the same output size, then you must have a larger sensor.  So, you’re saying there is no benefit in the Sony 50c sensor over the Canon 5DsR?  From memory, the sensors were released at about the same time …. From experience, the larger sensor are has larger sensel sites with significant benefits in image quality …

Yes, I think we’d worked out that all 60 million (more or less) pixels are read, and therefore Leica’s reference to “pixel binning” was shorthand for clever in camera processing.  Apparently, better than what LR can achieve …