Mavic Pro 2 - image sometimes soft ???

[parkgt]

So what is the take away from all this? Other posters suggest that the line skipping on FOV makes it worthless.

It seems like HQ D-log is the least problematic way to film right now?

 

[Roamer105]

HQ DlogM is my choice for sure. I lightly push sharpness and contrast (Landscape style setting) to overcome the "lazy encoder" issue (which is causing the mushiness). I use the free Ground Control LUT for M2P DlogM as a base, then fine-tune the grade as needed.

Also, unfortunately I don't think the test pattern I was "sold as calibrated for UHD when printed to 11" by 17" is a good test. It just does not look right even when printed to specs. But what is rather interesting in my last reply is comparing the picture taken by the Canon D7 Mark II with the professional (L lens) $1500 lens to the picture taken as a frame grab from HQ UHD video (Normal mode, Style=none) with the Mavic 2 Pro. Quite revealing!

 

[gnirtS]

Another issue with these tests is the drone optics arent optimised for ultra close up shots such as this (as one of the hasselblad designers states).
Not all monitors are equal either. Some look "nice" but in fact have slow, uneven response, aren't colour profiled and calibrated and so on. All TVs are generally useless for optical tests for those reasons too.

"Standard" certainly applies a fair bit of sharpening at 0 setting vs DLOG as well. It's a very very obvious difference.

 

[Cliff_622]

A lot of people confuse "sharpness" and "detail". The two actually have NOTHING to do with each other.

"Sharpness" only amounts to "micro contrast". It can also put a "halo" around pixels to make edges "pop". Yes, "sharpness" can give the illusion that you are seeing more detail and more resolution but you are NOT actually seeing any additional detail whatsoever. It's only making your brain see bolder edges. "Sharpness" adds ZERO new information into your image. You can creatively sharpen a DVD a million different ways and it can look SUPER DUPER CRAZY "sharp"... but it will NEVER have the actual "resolution" of a Blu-ray.

Actual image detail and resolution are more about higher amounts of finer "information" in your image. Pixels and lines getting closer together. It ultimately has to do with an image's higher "frequency" sampling. It's sounds goofy but...think of detail as a "low frequency" image vs a "high frequency" image. Micro-contrast ("sharpness") does not change the frequency of the image. Real "Detail" represents a higher frequency image, not just bolder contrast around existing edges.

Anyhoo,...that is too deep for a Saturday morning! I'm just saying that we often greatly confuse "sharpness" with "resolution" all the time....even though they are two completely different subjects!

On Ambarella processors, we use +1 sharpness, not because we want a "sharper" image! No no no!,....we do it so that it (luckily for us) Ambarella eases up on those terrible "detail" destruction and "smudging" and "smearing" artifacts that it's noise reduction processing does to the image. Remember once you wipe detail from your image,....i.e. lower it's frequency, it's gone forever and cannot be recovered...period.

CT

 

[Lolly]

I just did a sharpness test on the M2P shooting stills from f5 through f9 (keeping ev balanced with shutter speed down to 1/100th at slowest) and refocusing with AF with each shot..... The image got softer with each stopping down of the lens confirming what I'd noticed that smaller apertures result in soft imagery.
Could be AF issue?

 

[Paul2660]

I just did a sharpness test on the M2P shooting stills from f5 through f9 (keeping ev balanced with shutter speed down to 1/100th at slowest) and refocusing with AF with each shot..... The image got softer with each stopping down of the lens confirming what I'd noticed that smaller apertures result in soft imagery.
Could be AF issue?

No, it's classic diffraction. Softness due to smaller aperture. As you stop down, you increase diffraction. Diffraction is also effected by pixel size. The pixels on the 20MP 1" sensor a pretty small, much smaller than say pixels on a AF-C 24MP or 35mm 46MP chip. So by F 6.3 you have an issue, and by F9, pretty harsh diffraction.

If you shoot in raw you can recover some of it, but the finer stuff is gone.

F 3.5 to F 5.6 is the sweet spot IMO. Past that on anything with a large hyperfocal distance, and the diffraction problem will be noticeable.

AF on the MP2 is also important as attempting to manually focus is next to impossible for subjects that are 1 mile or even 1/2 mile off, at least with an iOS device.


Paul C

 

[CanadaDrone]

So what is the take away from all this? Other posters suggest that the line skipping on FOV makes it worthless.

It seems like HQ D-log is the least problematic way to film right now?

HQ D-log is the best, but the only issue with that is you need to manually correct for the fairly heavy barrel distortion in post. For the people using less advanced software, I am not sure how easy that is for them to do.

 

[CanadaDrone]

I just did a sharpness test on the M2P shooting stills from f5 through f9 (keeping ev balanced with shutter speed down to 1/100th at slowest) and refocusing with AF with each shot..... The image got softer with each stopping down of the lens confirming what I'd noticed that smaller apertures result in soft imagery.
Could be AF issue?

No AF issue there. The M2P (or more accurately the 1" Sony sensor inside) will start to have it's image degrade after F4 due to diffraction. By F11 it is smearing detail intended for a single pixel over approximately 21 pixels, which is why it looks so bad at tiny apertures.

My more detailed explanation here:
Mavic Pro 2 Image Quality

 

[Roamer105]

HQ D-log is the best, but the only issue with that is you need to manually correct for the fairly heavy barrel distortion in post. For the people using less advanced software, I am not sure how easy that is for them to do.

Very easy for any post software capable to grade Dlog.

 

[Paul2660]

A lot of people confuse "sharpness" and "detail". The two actually have NOTHING to do with each other.

"Sharpness" only amounts to "micro contrast". It can also put a "halo" around pixels to make edges "pop". Yes, "sharpness" can give the illusion that you are seeing more detail and more resolution but you are NOT actually seeing any additional detail whatsoever. It's only making your brain see bolder edges. "Sharpness" adds ZERO new information into your image. You can creatively sharpen a DVD a million different ways and it can look SUPER DUPER CRAZY "sharp"... but it will NEVER have the actual "resolution" of a Blu-ray.

Actual image detail and resolution are more about higher amounts of finer "information" in your image. Pixels and lines getting closer together. It ultimately has to do with an image's higher "frequency" sampling. It's sounds goofy but...think of detail as a "low frequency" image vs a "high frequency" image. Micro-contrast ("sharpness") does not change the frequency of the image. Real "Detail" represents a higher frequency image, not just bolder contrast around existing edges.

Anyhoo,...that is too deep for a Saturday morning! I'm just saying that we often greatly confuse "sharpness" with "resolution" all the time....even though they are two completely different subjects!

On Ambarella processors, we use +1 sharpness, not because we want a "sharper" image! No no no!,....we do it so that it (luckily for us) Ambarella eases up on those terrible "detail" destruction and "smudging" and "smearing" artifacts that it's noise reduction processing does to the image. Remember once you wipe detail from your image,....i.e. lower it's frequency, it's gone forever and cannot be recovered...period.

CT

These are good points, often overlooked. However sharping is still a needed requirement since all Bayer pattern sensors only capture 1 color, red, green, or blue. So if you capture red with a pixel, then you have to interpolate the green and blue. Anytime you interpolate you are guessing, and induce softness. This is true with any Bayer pattern sensor. Only a Foveon where each color is captured on a separate sensor can claim no interpolation for color.

Each jpg algorithm is adding some sharpening even if you turn off sharpening, it has to when it interpolates the colors.

Raw is no different again only one color is captured, but at least you can control a lot more in the post processing of the image.

Paul C

 

[Roamer105]

These are good points, often overlooked. However sharping is still a needed requirement since all Bayer pattern sensors only capture 1 color, red, green, or blue. So if you capture red with a pixel, then you have to interpolate the green and blue. Anytime you interpolate you are guessing, and induce softness. This is true with any Bayer pattern sensor. Only a Foveon where each color is captured on a separate sensor can claim no interpolation for color.

Each jpg algorithm is adding some sharpening even if you turn off sharpening, it has to when it interpolates the colors.

Raw is no different again only one color is captured, but at least you can control a lot more in the post processing of the image.

Paul C

Sharpness has very little to do with color capture. 4-2-0 color capture is only 1/4 resolution anyway and the human eye has far fewer color receptors than black/white. So your argument holds only true for color quality, but not image sharpness.

 

[Paul2660]

Interpolation for color equals loss of detail. It’s been written about for years and the main reason for chips like Foveon.

Interpolation of anything means loss of detail. True in uprez of an image also.

Discussion probably outside the scope of this post anyway.

Everyone has their own workflow for a digital image mine always includes some amount of sharpening depending final output printed or screen for web.

Paul C

 

[CanadaDrone]

Very easy for any post software capable to grade Dlog.

Grading yes - but what about barrel distortion correction? I am not even sure if LumaFusion can do it - probably, but I haven't checked.

 

[SteveCA]

In my tests M2P video looks softer almost on all color profiles vs original Mavic Pro, which I used to set sharpness to +1. I am not pro photographer and prefer more crisp video. Interesting if you look at single frame in editor, they look somehow sharper than over all exported video. Also I noticed increasing sharpness on M2P doesn't give you so much visible picture change as MP. (need more extensive testing in different lights) I don't want to say it is bad picture on M2P, in my opinion it is great, just different. It is different camera, so picture may vary as well.
Like I mentioned, I am not a pro, but noticed that better sharpness on M2P can be achieved also by adjusting aperture settings to get different depth of field results. Unfortunately weather in my area doesn't cooperate lately, so I wasn't able to fly/test a lot. If someone can share some insights on M2P aperture / shutter / sharpens settings, that would be very helpful.
I know there are some concerns about M2P, but IMO it is great drone and it needs to mature with future fw updates.
The only real problem I see with M2P now, it makes my beloved MP only as so, so drone....

I’ve done fairly extensive testing of the aperture settings and can conclude that stopping down is not the way to go with this lens. Imho this is a cheap, low quality lens, but to be expected at this price point. It is only sharp from F4.0-5.6. F8 and higher is total mush. F2.8 is soft but not as bad as F8-11. I only shoot at 4.0-5.0 during the day and will use 2.8 only at night. my videos are sharp using +1. FOV is sharp but not as sharp as HQ.

Try optimizing aperture as it has a greater effect on detail than the difference between FOV and HQ IMO.

I am an avid landscape photographer and have been testing lenses for optimal apertures for many years. Only the pro grade glass, which I use exclusively, are really sharp fully stopped down. all lenses have a sweet spot for sharpness. The most expensive ones are tack sharp from F2.8 all the way to F11 but we are talking thousands of dollars in most cases.

 

[CanadaDrone]

I’ve done fairly extensive testing of the aperture settings and can conclude that stopping down is not the way to go with this lens. Imho this is a cheap, low quality lens, but to be expected at this price point. It is only sharp from F4.0-5.6. F8 and higher is total mush. F2.8 is soft but not as bad as F8-11. I only shoot at 4.0-5.0 during the day and will use 2.8 only at night. my videos are sharp using +1. FOV is sharp but not as sharp as HQ.

Try optimizing aperture as it has a greater effect on detail than the difference between FOV and HQ IMO.

I am an avid landscape photographer and have been testing lenses for optimal apertures for many years. Only the pro grade glass, which I use exclusively, are really sharp fully stopped down. all lenses have a sweet spot for sharpness. The most expensive ones are tack sharp from F2.8 all the way to F11 but we are talking thousands of dollars in most cases.

On the M2P this is primarily due to diffraction which is completely independent of lens quality. It is something that can even be calculated in advance without even owning the drone - explanation here:

Mavic Pro 2 Image Quality

The DSLRs you (presumably) use for landscape photography are bound by the exact same rules of physics, it's just harder to notice with bigger sensors and large pixel size.

Tested in isolation, most DSLR lenses have MTF peaks around F4-F5.6 except for a very select few that are designed to be sharpest at or near wide open. Typical behavior is as you approach F8-F11, center and overall sharpness is traded for slightly better corners. This is something optical engineers can play with, but everything is a trade-off. On most FF cameras, much beyond F11, approaching F16+, diffraction is getting so bad that it doesn't matter how good the lens is, it will look very soft at the pixel level. This is still separate from diffraction and can be measured without even having a camera in the equation.

 

[SteveCA]

On the M2P this is primarily due to diffraction which is completely independent of lens quality. It is something that can even be calculated in advance without even owning the drone - explanation here:

Mavic Pro 2 Image Quality

The DSLRs you (presumably) use for landscape photography are bound by the exact same rules of physics, it's just harder to notice with bigger sensors and large pixel size.

Tested in isolation, most DSLR lenses have MTF peaks around F4-F5.6 except for a very select few that are designed to be sharpest at or near wide open. Typical behavior is as you approach F8-F11, center and overall sharpness is traded for slightly better corners. This is something optical engineers can play with, but everything is a trade-off. On most FF cameras, much beyond F11, approaching F16+, diffraction is getting so bad that it doesn't matter how good the lens is, it will look very soft at the pixel level. This is still separate from diffraction and can be measured without even having a camera in the equation.

Thanks for the detailed explanation of why the Mavic 2 Pro lens is soft at apertures other than 4.0 or thereabouts. While I dont profess to understand the science behind this, I know what apertures to use if detail and sharpness are important. That's why I only shoot at 4.0 to 4.5 during the day, using ND filters to accomplish that while maintaining a 2x shutter speed for video.