Mosaic pattern on DNG raw photos

[Jfvdenning]

I have been looking at DPreview dog raw files from The Mavic 3 as I use drone for photography and there is some mosaic pattern in the raw files, I can see a pattern when zoomed in all over the images, it is most obvious on skies as shown in my screenshot. It is not visible in jpegs which I guess means it is a software issue but as these are DNG files they should be readable by Lightroom with out issue.

Here at the image samples - DJI Mavic 3 Cine sample gallery

 

[eEridani]

I imported one of the DNG's (025) and it looks fine in ACR/CS5.

 

[Meta4]

I have been looking at DPreview dog raw files from The Mavic 3 as I use drone for photography and there is some mosaic pattern in the raw files, I can see a pattern when zoomed in all over the images, it is most obvious on skies as shown in my screenshot. It is not visible in jpegs which I guess means it is a software issue but as these are DNG files they should be readable by Lightroom with out issue.

Here at the image samples - DJI Mavic 3 Cine sample gallery

You've enlarged that part of the image about 300%.
Do the same to the jpg and it looks like mush.

But if you enlarge that DNG further, you can see that what you called a mosaic pattern is just the individual pixels.

What you are seeing is just the way the raw image viewer displays over-enlargement.
It's not a problem with the image file.

Try doing the same to dng files from one of your own cameras and see how that looks.

 

[Jfvdenning]

You've enlarged that part of the image about 300%.
Do the same to the jpg and it looks like mush.

But if you enlarge that DNG further, you can see that what you called a mosaic pattern is just the individual pixels.

What you are seeing is just the way the raw image viewer displays over-enlargement.
It's not a problem with the image file.

Try doing the same to dng files from one of your own cameras and see how that looks.

No actually I did not, I am aware of that effect when over enlarging on photoshop and it requires more than 300 percent to see it, this is Lightroom which does not do that when enlarging. This is different as you can obviously see looking at my screenshot as the pattern changes.
It seems to be an issue with how Lightroom has done the demosaicing, and could be from this being a Quad Bayer sensor and it not interpreting it correctly, and eventual update with most lily fix it.

 

[Meta4]

No actually I did not, I am aware of that effect when over enlarging on photoshop and it requires more than 300 percent to see it, this is Lightroom which does not do that when enlarging. This is different as you can obviously see looking at my screenshot as the pattern changes.
It seems to be an issue with how Lightroom has done the demosaicing, and could be from this being a Quad Bayer sensor and it not interpreting it correctly, and eventual update with most lily fix it.

When I enlarged the same image 300% in Photoshop, it looked exactly like what you posted.
When I enlarged it 1200%, it looks like what I posted.

 

[adcimagery]

Guessing Lightroom needs an update for these raw files for best compatibility. Could be a debayering/demosaicing issue, or related to the phase detect pixels on the sensor (if present).

 

[Meta4]

Guessing Lightroom needs an update for these raw files for best compatibility. Could be a debayering/demosaicing issue, or related to the phase detect pixels on the sensor (if present).

There's nothing wrong with the image or Lightroom.
That's just how it looks when enlarged beyond 100%

 

[adcimagery]

There's nothing wrong with the image or Lightroom.
That's just how it looks when enlarged beyond 100%

At 100%, clear banding is visible. This is not some sort of interpolation issue from zooming in.

Here's an example with some sharpening added. The black lines illustrate the gaps in the banding. This is an artifact of raw processing, a banding problem inherent to the sensor design, or a firmware issue related to something like sensor readout, amplification, or ADC (I'm assuming the press review drones were running pre production firmware).

Looking at the image closely, I'm going to guess it's just really bad demosaicing - there's a ton of false color/moire issues with fine patterns. Particularly egregious is the Chase bank logo in that middle circle.

 

[Jfvdenning]

When I enlarged the same image 300% in Photoshop, it looked exactly like what you posted.
When I enlarged it 1200%, it looks like what I posted.

And you can't see how completely different they look? Go enlarge it 1200 percent in Lightroom and you will see that is just how photoshop displays all pixels at that scale. That is not the issue here.

 

[Jfvdenning]

There's nothing wrong with the image or Lightroom.
That's just how it looks when enlarged beyond 100%

Please stop replying as you do not know what you are talking about. Thanks for trying though but there is something wrong.

 

[Jfvdenning]

At 100%, clear banding is visible. This is not some sort of interpolation issue from zooming in.

Here's an example with some sharpening added. The black lines illustrate the gaps in the banding. This is an artifact of raw processing, a banding problem inherent to the sensor design, or a firmware issue related to something like sensor readout, amplification, or ADC (I'm assuming the press review drones were running pre production firmware).

Looking at the image closely, I'm going to guess it's just really bad demosaicing - there's a ton of false color/moire issues with fine patterns. Particularly egregious is the Chase bank logo in that middle circle.

The more I look at the pattern the more it looks like the sensor is capturing the image at different exposures across the sub pixels underneath the RGB filter and Lightroom doesn't know how to read the data correctly yet.

 

[adcimagery]

The more I look at the pattern the more it looks like the sensor is capturing the image at different exposures across the sub pixels underneath the RGB filter and Lightroom doesn't know how to read the data correctly yet.

Yeah, lot of links in the image chain means it could be in the sensor, drone firmware, or raw processor. I’d doubt it’s an intractable problem, though - probably a Lightroom update will fix it, assuming it’s not a drone f/w issue. Curious to see if one of the dedicated raw analysis tools will show it.

 

[Scotdrone]

At 100%, clear banding is visible. This is not some sort of interpolation issue from zooming in.

Here's an example with some sharpening added. The black lines illustrate the gaps in the banding. This is an artifact of raw processing, a banding problem inherent to the sensor design, or a firmware issue related to something like sensor readout, amplification, or ADC (I'm assuming the press review drones were running pre production firmware).

Looking at the image closely, I'm going to guess it's just really bad demosaicing - there's a ton of false color/moire issues with fine patterns. Particularly egregious is the Chase bank logo in that middle circle.

I've been working with sample DNG files in advance of committing to purchasing the Mavic3 and the false colour/moire issue was one that I spotted on all the example files I downloaded. It doesn't seem to appear on the camera generated jpeg so the drone must have some kind of correction filter built in to the way it generates the jpegs, but currently if you're working with DNGs in Lightroom etc, eliminating the distracting colour fringing takes a lot of time and effort. I'm not sure where the solution to this lies.

 

[BingBong]

Mavic 3 is not on Adobe's supported Raw cameras list yet, so I suspect an update will solve the problem.

https://helpx.adobe.com/camera-raw/kb/camera-raw-plug-supported-cameras.html#DJI

New proprietary Raw formats are not usually readable by Camera Raw until Adobe supports them. People sometimes force it by spoofing an earlier model, but that will obviously not have a profile for the new model, so may perform poorly. DNG being a generic format means that it will always be readable immediately, but that doesn't imply support is there. An update will be needed.

 

[adcimagery]

I've been working with sample DNG files in advance of committing to purchasing the Mavic3 and the false colour/moire issue was one that I spotted on all the example files I downloaded. It doesn't seem to appear on the camera generated jpeg so the drone must have some kind of correction filter built in to the way it generates the jpegs, but currently if you're working with DNGs in Lightroom etc, eliminating the distracting colour fringing takes a lot of time and effort. I'm not sure where the solution to this lies.

Interesting. I'm looking at a production version's files running the latest firmware, in the most recent update of Lightroom Classic, and I'm not seeing the issue. Wondering if it's firmware related as the latest update mentioned image quality improvements.

 

[sar104]

I've been working with sample DNG files in advance of committing to purchasing the Mavic3 and the false colour/moire issue was one that I spotted on all the example files I downloaded. It doesn't seem to appear on the camera generated jpeg so the drone must have some kind of correction filter built in to the way it generates the jpegs, but currently if you're working with DNGs in Lightroom etc, eliminating the distracting colour fringing takes a lot of time and effort. I'm not sure where the solution to this lies.

Can you provide a link to any examples of the problem?

 

[adcimagery]

Can you provide a link to any examples of the problem?

I see it in that linked gallery from dpreview, but not in the images I just captured in the aforementioned conditions. Lightroom's default definitely has room for improvement, but it's much better than the Dpreview examples when it comes to fine detail.

 

[Jfvdenning]

I see it in that linked gallery from dpreview, but not in the images I just captured in the aforementioned conditions. Lightroom's default definitely has room for improvement, but it's much better than the Dpreview examples when it comes to fine detail.

Interesting, it seems like it could be the firmware then. DNG is universal so even with a new camera old software should be able to read it. My old Lightroom even has the correct lens corrections for the lens on this Mavic 3, and states the correct model L2D-20C, as it is built into the DNG spec so no update is needed.

 

[Johnmcl7]

Interesting, it seems like it could be the firmware then. DNG is universal so even with a new camera old software should be able to read it. My old Lightroom even has the correct lens corrections for the lens on this Mavic 3, and states the correct model L2D-20C, as it is built into the DNG spec so no update is needed.

The Mavic drones are the only time I've worked with DNG files (all my cameras produce proprietary raw files) but I was a bit puzzled how software would need an update to work with a standard file type. I understand why raw converters need updates for new raw files when they're proprietary since each is different but didn't think the same should apply to DNG files.

 

[BingBong]

The Mavic drones are the only time I've worked with DNG files (all my cameras produce proprietary raw files) but I was a bit puzzled how software would need an update to work with a standard file type. I understand why raw converters need updates for new raw files when they're proprietary since each is different but didn't think the same should apply to DNG files.

Same for me, but the M2P and A2S have specific profiles listed by Adobe for Camera Raw. I think Adobe add their tweaks, which could well improve moiré and many other factors in subtle ways.