Leica SL3 Firmware Update / Bug Fix

[Stuart Richardson]

Advertisement (gone after registration)

I don’t know how exactly they stitch it all together, but it is immediately apparent that the pixel dimensions and captured detail are higher. I compared with taking the original file and using super resolution in Adobe and the multishot had more detail resolution than the interpolated file, so it is not just interpolation. That said, I don’t believe it is equivalent to a true 187mp. I agree with Gordon that it is more like 100 or so. 

[SrMi]

2 hours ago, Stuart Richardson said:

I don’t know how exactly they stitch it all together, but it is immediately apparent that the pixel dimensions and captured detail are higher. I compared with taking the original file and using super resolution in Adobe and the multishot had more detail resolution than the interpolated file, so it is not just interpolation. That said, I don’t believe it is equivalent to a true 187mp. I agree with Gordon that it is more like 100 or so. 

Also note that the motion artifact removal option uses a blown-up single-shot image to replace the artifacts, i.e., the results at those patches are lower than what you get in the post.

[jaapv]

2 hours ago, Stuart Richardson said:

I don’t know how exactly they stitch it all together, but it is immediately apparent that the pixel dimensions and captured detail are higher. I compared with taking the original file and using super resolution in Adobe and the multishot had more detail resolution than the interpolated file, so it is not just interpolation. That said, I don’t believe it is equivalent to a true 187mp. I agree with Gordon that it is more like 100 or so. 

And that is because the uniform areas are disregarded in the stitching software pipeline.  The net effect is virtually identical. Compare to lossless compression.

[FlashGordonPhotography]

I wasn’t making my answer up. Really. Feel free to do your own research as usual but I have on this one. Jim Kasson and others etc. Read some papers on Foveon sensors which is very interesting tech and how actual data integrity affects the RGB file we work with.

My best understanding is. All bayer files are interpolated. For a single shot 2/3 of the information is made up. It’s also a requirement of the RGB system that we have a value for RGB at each *pixel*, currently in post software. So our SL2, 47 million photosite capture ends up as a 47 x 3 pixel file as the camera interpolates the two extra values at each photosite. So far so good.

Normal high res shot is a 4 capture process. One at the original location and three additional shots shifted one pixel in three directions. This means you get each photosite to capture each colour correctly at that shotosite. Now you have a 47 x 4 x 3 pixel file. 188MP file (rounded down to 187 for the edge pixels). Four captures in a RGB file. But each shift is ONE photosite. You don’t capture more information but you do capture more accurate information. You lower the noise floor and *may* improve DR.Technically you’re not capturing more information but you are capturing better quality (more accurate) detail. You’ve given the RGB system more data to work with so the end file is more accurate and less prone to false information. It’s more like high res shot revealed more detail that was hidden by false information than actually capturing more detail. There are also 8 and 16 shot multi shot systems and these do send more actual detail to the system. AFAIK the SL2 uses 4 shot. There’s also half pixel shift systems etc.

Super resolution and Gigapixel etc just interpolate already interpolated data. Mostly this works well. But it’s not actual data and sometimes there’s a difference because colour aliasing hides detail and upressing just works with *faulty* data. As we move into a world of more powerful AI this difference will shrink but for now high res shot will show more detail than an unpressed file usually. So far, I’m finding that AI NR like Lightroom and DXO before just using a bicubic upressing in Photoshop or Lightroom’s print module delivers as good or better files than super resolution. I think this is because as part of that AI process the software finds aliasing and replaces it. But I’m not absolutely sure. A time will come when software does make it indistinguishable. That time is not yet.

This process is exactly the same reason the a Monochrom sensor, properly exposed and filtered, will show more detail than an equivalent RGB file. Actual data vs interpolated data. High res shot is like shooting the Monochrom camera and upressing is like using a colour file to make a monochrom image.

I am not dismissing the use of usefulness of upresssed files using modern software. Modern upressing software makes excellent files. But high res shot makes better files and actually increasing the pixel count makes better files again. You may not see it until you get to the bleeding edge but it’s there. Otherwise we could have stopped at 11MP cameras.

But as always, please do your own research. I’m not a physicist. I just imitate one sometimes.

Gordon

[Virob]

Multishot on the SL2 combines 8 captures and the shift is 1/2 pixel. This results in better color fidelity and increased resolution.

[Photoworks]

Since the update, the camera is always been getting hot, especially the park of the EVF by the big LEICA name.

Is anyone else getting this?

[FlashGordonPhotography]

Advertisement (gone after registration)

1 hour ago, Photoworks said:

Since the update, the camera is always been getting hot, especially the park of the EVF by the big LEICA name.

Is anyone else getting this?

No but I’ve only used the camera sporadically since the update. Happy to try and replicate. How long is this taking? Photo or video? And special use case like AFC etc?

Gordon

[Photoworks]

7 hours ago, FlashGordonPhotography said:

No but I’ve only used the camera sporadically since the update. Happy to try and replicate. How long is this taking? Photo or video? And special use case like AFC etc?

Gordon

few hours USB tethered in MF on a tripod. Bracket exposure and 24-90 zoom. thanks

[jaapv]

On 7/20/2024 at 12:39 PM, SrMi said:

Because multishot increases the real resolution only by a little, mainly by removing any false colors (moire). Moire is is more of a problem with 24MP than with 60MP sensors, especially when no AA Filter is used.

It is just the other way around. The increase in resolution reduces colour aliasing and moire. 

[SrMi]

6 minutes ago, jaapv said:

It is just the other way around. The increase in resolution reduces colour aliasing and moire. 

Yes for real resolution, not for multishot “resolution.”

[jaapv]

Multishot resolution shifts pixels. As such it is identical as it adds pixels to the image. . The difference is that it uses some of these pixels to correct artifacts -that is why it does not take one shifted image but a series, and that it does not multiply the pixel count in uniform areas.  Thus the observed pixel count will always be less than a native high resolution image but the effect of increased detail will be present And the fineness of the detail will reduce aliasing and moire. 
I find the effect of high resolution images to be quite overblown. Moire and aliasing As a nature photographer shooting amongst other things a fair number of birds it should be a problem. But at 24 MP it is a rare occurrence. And increased detail? No viewer has ever known that some finest detail is missing as he never knew it was there. What matters is the quality and acuity of the detail and that is where the newest sensors -including the S5ii shine due to the increased potential of the latest EUV lithography machines. A modern 24 MP sensor will yield superior results to a first generation high resolution sensor. Croppability? I prefer a longer lens 😉

[FlashGordonPhotography]

4 hours ago, SrMi said:

Yes for real resolution, not for multishot “resolution.”

To the best of my knowledge multi shot is very effective with moire and other colour issues. The whole point is to get actual colour data to eliminate these things. That’s what high res shot does. 

4 hours ago, jaapv said:

Multishot resolution shifts pixels. As such it is identical as it adds pixels to the image. . The difference is that it uses some of these pixels to correct artifacts -that is why it does not take one shifted image but a series, and that it does not multiply the pixel count in uniform areas.  Thus the observed pixel count will always be less than a native high resolution image but the effect of increased detail will be present And the fineness of the detail will reduce aliasing and moire. 
I find the effect of high resolution images to be quite overblown. Moire and aliasing As a nature photographer shooting amongst other things a fair number of birds it should be a problem. But at 24 MP it is a rare occurrence. And increased detail? No viewer has ever known that some finest detail is missing as he never knew it was there. What matters is the quality and acuity of the detail and that is where the newest sensors -including the S5ii shine due to the increased potential of the latest EUV lithography machines. A modern 24 MP sensor will yield superior results to a first generation high resolution sensor. Croppability? I prefer a longer lens 😉

Oh, they notice. As someone who now specialises in large fine art prints we have data. Even at conservative sizes shots from cameras over 40MP outsell images from 24MP by a significant margin (nearly 4:1). We don’t print lower res files over A1 so the data isn’t possible for larger than that but it’s a stark difference. The difference between 40 and 100MP is much smaller. Our most popular seller is from a 42MP camera and it skews the data a bit because it really does sell well. Photographers go on about this stuff but buyers like sharpness and detail. Nick Brandt showed this is also true for wildlife. Ken Duncan, an Australian 617 landscape photographer has his gallery about 5 minutes from me so I realised this a long long time ago.

This doesn’t mean lower resolution files can’t make staggering images. But if you could have side by side with different resolutions, high res wins. Mind you, you need the lenses first. A lower resolution image with great glass can have more real world resolution than a high res camera with average glass. I think my S5II would make better files with the 60-600 than my SL3 with 150-600 (my dodgy copy). Mind you I just put the 60-600 or a 500 prime on the SL3 and be done with it.

I prefer to crop with a lens as well. Or my feet if it’s possible.

Gordon

[SrMi]

3 hours ago, FlashGordonPhotography said:

To the best of my knowledge multi shot is very effective with moire and other colour issues. The whole point is to get actual colour data to eliminate these things. That’s what high res shot does. 

Yes. A four-shot multi-shot will reduce moiré and produce output with the same dimensions (resolution in pixels). An eight or higher shot multishot also increases the output dimension (resolution in pixels). Moiré is also reduced by using sensors with higher resolution.

[jaapv]

And eliminated in post. A new member has offered a Photoshop action that is quite effective.  And saves a lot of work. 

 

[jaapv]

11 hours ago, SrMi said:

Yes. A four-shot multi-shot will reduce moiré and produce output with the same dimensions (resolution in pixels). An eight or higher shot multishot also increases the output dimension (resolution in pixels). Moiré is also reduced by using sensors with higher resolution.

Yes. Panasonic uses at least ten shots. The resulting resolution is 12000x8000

[SrMi]

4 hours ago, jaapv said:

Yes. Panasonic uses at least ten shots. The resulting resolution is 12000x8000

Maybe that is for handheld multishot, which is different from pixel shift. Tripod pixel shift is typically implemented as an 8, 16, or 32-shot implementation. I read that S5II uses eight shots (link), like S1R. Do you have sources that say otherwise?

[jaapv]

No - I count as far as possible at that speed.

[SrMi]

23 hours ago, jaapv said:

No - I count as far as possible at that speed.

Tripod or handheld? Handheld mode is not really pixel shift and takes more images, some of which get rejected. I was talking about tripod mode (pixel shift).

[Photoworks]

On 7/24/2024 at 12:01 PM, jaapv said:

And eliminated in post. A new member has offered a Photoshop action that is quite effective.  And saves a lot of work. 

 

most action in Photoshop relies on blurring channels. it takes away the issue, but a noticeable loss in resolution.

[Advertisement]

Hallo,

Look here: Leica L-Mount

[jaapv]

Possibly- I have to test this one yet.