Poor image quality

[erhard]

All good. I give up.

Yes, I understand that.

That, to me, is reason to get rid of it, and use the (in your judgement) superior mini 2.



Simply this: What seems most important to you here is getting others who judge this differently than you to admit they're wrong. You can't accept, "it looks different to me than it does to you".

Now at 140 posts, it's pretty clear how this subjective issue is subjectively being judged, opinions aren't changing, so it's getting hard to see what is being accomplished by posting images, saying they're blurry, and having the same people come back again and again and say they're not.

 

[Occams Razor]

ok here now side by side comparison which one is softer?
here the thumbnail which shows you the full image so you can see the section the 100% crops are taken out. so the right side of the crops are about at center image

View attachment 177221

and here the side by side - drone 15m up. sure we can blame out of focus but then all pics from m4p are out of focus and that on 3 different f1.7 lens drones 1xm3p 2xm4p - very unlikely. we can argue what would be the reason why one is softer but lets just agree that one is softer - you tell me which one please. thanks for sticking with me for so long .
you or someone else suggested i should sell. one point why this is a hard thing to do is today it was very windy and the mini2 gave me a lot of strong wind warning whereas the mini4p behaved much more relaxed under the same conditions. as i said it is a great drone but not great image quality . i did not add my f2.8 aperture on the mini4 to null the argument of focus error but i did this in previous pics and saw no difference.
hope that is a good enough argument.
also the m4p pics where taken in in pro mode which per manual (and DJI confirmed) states that no HDR processing was being applied. i did this because it was brought up in discussions that HDR for example on the iphone reduces the resolution.

View attachment 177222

The image on the right was processed by a JPEG engine that is more aggressive than that on the left. The image on the right is over-sharpened, has greater saturation and the contrast or black point is much stronger as compared to the left photo. Can you post the DNGs and JPEGs for both images on your One-Drive?

 

[tuxtard]

This still seems like just a difference in sharpening settings to me. I’m not sure why there are no sharpening settings in photo mode on the M4P, while they are available on the M3P. Sharpening settings should be included so that everyone can adjust them to their preference.

 

[erhard]

This still seems like just a difference in sharpening settings to me. I’m not sure why there are no sharpening settings in photo mode on the M4P, while they are available on the M3P. Sharpening settings should be included so that everyone can adjust them to their preference.

I tried to edit the raw in a raw editor but was unable to fix it but I also admit I have no experience to work with raw files. I also could not improve the jpeg.
At the time when I had the m3p for a few days only and got the same issue I did not realize it had the saturation setting available. Maybe someone with m3p can play with it and show the effect.

 

[erhard]

The image on the right was processed by a JPEG engine that is more aggressive than that on the left. The image on the right is over-sharpened, has greater saturation and the contrast or black point is much stronger as compared to the left photo. Can you post the DNGs and JPEGs for both images on your One-Drive?

ok posted here
DJI mini2 mini4p compare

the 2 filenames of interest are these but the DNG versions - the 0081 is the mini2

 

[Occams Razor]

great, can you please attach a 4p image.
as i said many time i have attached to original pics unaltered right out of the SDcard- what more can i do, each about 5MB?

Thanks for posting the images. I took both DNG's and set all of the sliders in Photoshop including sharpening to zero. I wanted to compare the 2 images with minimal post processing. The Mini 2 is on the bottom. This is close to a full pixel view but the M4P image was a little closer to the track so that one is about 90%. The Mini 2 image is brighter but that is mainly because the M4P was set with exposure compensation at -.3. The M4P image also has a color cast which is interesting because I set the temperature to 4950 to try to get them as close as possible.

The unprocessed photos are very close. The Quad Bayer sensor from the M4P may produce DNG images with greater photographic latitude so that may be why people feel the photos are better in the end. I am leaving for a trip to Australia this evening so I don't have time to work with the photos more.

 

[erhard]

Thanks for posting the images. I took both DNG's and set all of the sliders in Photoshop including sharpening to zero. I wanted to compare the 2 images with minimal post processing. The Mini 2 is on the bottom. This is close to a full pixel view but the M4P image was a little closer to the track so that one is about 90%. The Mini 2 image is brighter but that is mainly because the M4P was set with exposure compensation at -.3. The M4P image also has a color cast which is interesting because I set the temperature to 4950 to try to get them as close as possible.

The unprocessed photos are very close. The Quad Bayer sensor from the M4P may produce DNG images with greater photographic latitude so that may be why people feel the photos are better in the end. I am leaving for a trip to Australia this evening so I don't have time to work with the photos more.

View attachment 177265

Thanks for your time and expertise. I will. Heck them out on the bigger screen when home

As I mentioned I have no experience with raw processing but used the raw therapy program to look at the Dng but the knobs are too many for me to adjust. I used the pro Mode so there is no HDR involved and all the quad Bauer should do is give a bigger sensor area per pixel I think.
The good news is that with the Dng I should be able to process the images later but I would be prefer to have jpegs to my liking so maybe DJI will add a few more adjustment capabilities to the fly app
Anyways thanks again and have a safe trip

Erhard

 

[erhard]

Thanks for posting the images. I took both DNG's and set all of the sliders in Photoshop including sharpening to zero. I wanted to compare the 2 images with minimal post processing. The Mini 2 is on the bottom. This is close to a full pixel view but the M4P image was a little closer to the track so that one is about 90%. The Mini 2 image is brighter but that is mainly because the M4P was set with exposure compensation at -.3. The M4P image also has a color cast which is interesting because I set the temperature to 4950 to try to get them as close as possible.

The unprocessed photos are very close. The Quad Bayer sensor from the M4P may produce DNG images with greater photographic latitude so that may be why people feel the photos are better in the end. I am leaving for a trip to Australia this evening so I don't have time to work with the photos more.

View attachment 177265

now this is interesting - the m4p and m2 look about the same but now the m2 is also soft. if i used the original m2 jpeg it is clearly sharper. just what i see, sorry

 

[Droning on and on...]

I used the pro Mode so there is no HDR involved and all the quad Bauer should do is give a bigger sensor area per pixel I think.

Nope. The Quad Bayer color filter pattern was invented by Sony to enable high dynamic range by binning pixels in a Bayer layout, and in a different mode capture 4x the pixels for more detail in QB mode.

When capturing images in 12MP mode, the sensor is functioning as a standard Bayer setup. In 48MP mode, the QB layout is used.

QB produces much more error in reproducing color for most pixels. Because of this, resolution is lost as a consequence of demosaicing. So you don't actually get 4x the resolution of 12MP mode. The qualitative increase in detail is around 1.5-2x.

 

[Occams Razor]

now this is interesting - the m4p and m2 look about the same but now the m2 is also soft. if i used the original m2 jpeg it is clearly sharper. just what i see, sorry

I removed sharpening from both images to show that it is the JPEG processing that is creating the main difference between images from the two drone cameras. i.e. the unprocessed RAW images are about the same.

 

[erhard]

I removed sharpening from both images to show that it is the JPEG processing that is creating the main difference between images from the two drone cameras. i.e. the unprocessed RAW images are about the same.

right, that is what was my hope. so if DJI is nice they would add some sliders for image adjustment in the future.

 

[erhard]

Nope. The Quad Bayer color filter pattern was invented by Sony to enable high dynamic range by binning pixels in a Bayer layout, and in a different mode capture 4x the pixels for more detail in QB mode.

When capturing images in 12MP mode, the sensor is functioning as a standard Bayer setup. In 48MP mode, the QB layout is used.

QB produces much more error in reproducing color for most pixels. Because of this, resolution is lost as a consequence of demosaicing. So you don't actually get 4x the resolution of 12MP mode. The qualitative increase in detail is around 1.5-2x.

that is true the 48MB does not really improve much so i will not use it very often. but DJI says HDR is used in AUTO mode even in the 12MB format and it is OFF in PRO mode which i used. my understanding is that the HDR QB is mainly used in video shooting. you say even when HDR is OFF in still images the QB is still behaving differently than just a larger pixel for lower noise?
so are you thinking that the m4p can produce the same resolution as the m2 jpeg with proper post processing on the DNG file?

 

[Meta4]

are you thinking that the m4p can produce the same resolution as the m2 jpeg with proper post processing on the DNG file?

I'm not sure what you mean by resolution.
In 12 MP mode, the max resolution of both cameras is the same .. 4000 x 3000.

 

[erhard]

I'm not sure what you mean by resolution.
In 12 MP mode, the max resolution of both cameras is the same .. 4000 x 3000.

right, but the mini2 jpeg is crisper

 

[erhard]

right, but the mini2 jpeg is crisper

that is what i mean the sensor has the same number of pixels just each one is larger in the m4p

 

[Meta4]

that is what i mean the sensor has the same number of pixels just each one is larger in the m4p

Since the sensor in the mini 4 pro is roughly twice the area of that in the Mini 2, it follows that the pixels will be larger for the Mini 4 (in 12MP mode).

 

[erhard]

Since the sensor in the mini 4 pro is roughly twice the area of that in the Mini 2, it follows that the pixels will be larger for the Mini 4 (in 12MP mode).

i think it is 1.7x. yes so it gives you better SNR and on top of it the f1.7 - in theory all good.
on my first m4p i had sadly terrible edge distortion, much better on the replaced one.

 

[Droning on and on...]

that is true the 48MB does not really improve much so i will not use it very often. but DJI says HDR is used in AUTO mode even in the 12MB format and it is OFF in PRO mode which i used. my understanding is that the HDR QB is mainly used in video shooting. you say even when HDR is OFF in still images the QB is still behaving differently than just a larger pixel for lower noise?
so are you thinking that the m4p can produce the same resolution as the m2 jpeg with proper post processing on the DNG file?

The quasi-HDR feature of the dual native ISO sensor is a completely different, independent feature of the sensor from the Quad Bayer filter layout.

The dual ISO feature is only available for video. Basically what it does is energize alternating rows of pixels with two different output amplifiers, getting the noise "sweet spot" at two different ISOs. For video resolution, rows of pixels on the 48MP sensor are already paired up anyway, so the data from the two photosites in the upper row, at native ISO 400, and the lower row at ISO 800 can be combined in a way similar to exposure bracketing to increase dynamic range.

This feature is not available for stills. It might make sense in 12MP mode, but generally if you're getting this into the weeds with stills you're better off – in my opinion – learning how to "develop" raw captures.

You're thread here makes that case, in my view.

 

[erhard]

The quasi-HDR feature of the dual native ISO sensor is a completely different, independent feature of the sensor from the Quad Bayer filter layout.

The dual ISO feature is only available for video. Basically what it does is energize alternating rows of pixels with two different output amplifiers, getting the noise "sweet spot" at two different ISOs. For video resolution, rows of pixels on the 48MP sensor are already paired up anyway, so the data from the two photosites in the upper row, at native ISO 400, and the lower row at ISO 800 can be combined in a way similar to exposure bracketing to increase dynamic range.

This feature is not available for stills. It might make sense in 12MP mode, but generally if you're getting this into the weeds with stills you're better off – in my opinion – learning how to "develop" raw captures.

You're thread here makes that case, in my view.

oh no, dark room again - i did this 60 years ago with my dad, ok this time it would be no chemicals involved.
i hope when you are you can teach my how to achieve a mini2 like look with mp4 dng. i can use rawtherapy hopefully to follow the recipe. curious, can photoshop generate an executable that allows non PS user to modify images?

 

[ExEvo]

M4P lens distortion would be evident in video too but I don't see any in this frame grab taken from a 4k video with the drone hovering 6ft above the gravel.

Dropbox

www.dropbox.com