48MP Bayer Sensor + Image Averaging = Increase in Signal to Noise

[Flashheart]

You can increase the signal to noise ratio of the 48mp quad bayer sensor on the M3P by employing a technique called image averaging. This technique corrects most of the noise and noticably improves the image quality for 48mp images.

The YouTube link below explains the technique for those who are interested.

To view this content we will need your consent to set third party cookies.
For more detailed information, see our cookies page.

Accept third party cookies

 

[CanadaDrone]

For anyone who may not be aware, that channel is one of the biggest jokes in the photographic community. They regularly make videos perpetuating myths and other very unhelpful things. Unfortunately, they are profiting on that controversy (that's what gets views these days) and that has only made things worse. They will say literally anything for likes & views, have become total sellouts based on whatever free gear they get sent by manufacturers, and have a bad reputation as a result. They switch camera systems every time they get sent something for free and make a big video about how X is better than Y based on whoever gave them free gear more recently. They regularly conduct comparison tests on gear they don't even understand how to use and try and make recommendations to people. Please don't give them any more views than they already get.

What he's doing there is nothing new or revolutionary, it's actually how most smartphones help improve their noise performance but they do it 'on the fly' without all the extra time in Photoshop. Noise occurs in all images randomly, so the more frames you stack/average, the lower the overall noise. It's a very basic principle that's been around for as long as digital photography and it works on any image, it's nothing specific to the Mini 3 Pro or any other camera or drone. It's also very tedious, and there is software these days that does a better job even with a simple batch process. Programs like DXO Pure RAW 2 or Topaz DeNoise give better results and with far less time/effort.

OP - I am not trying to rain on your parade, especially as there are likely a lot of folks not aware of that channel's reputation, but with how much blatant misinformation that channel spews in general, I would hate for people to watch their content and think they are a trustworthy source of information.

 

[Cymru]

In addition to that, any moving objects cause problems with averaging. There are usually plenty of things moving in aerial images.

 

[Flashheart]

In addition to that, any moving objects cause problems with averaging. There are usually plenty of things moving in aerial images.

It’s easy to mask out any ghosting. The pros outweigh the cons. IA also removes a lot of the artifacts that are inherent with the MP3’s quad Bayer sensor.

 

[Cymru]

Not easy to mask out water ripples, grass, trees and everything else. Youd do so much the entire image would be masked and unchanged.
Not to mention horrifically time consuming for every single image.

Not a practical or useful technique to correct low quality drone images.

 

[Flashheart]

Not easy to mask out water ripples, grass, trees and everything else. Youd do so much the entire image would be masked and unchanged.
Not to mention horrifically time consuming for every single image.

Not a practical or useful technique to correct low quality drone images.

Low quality drone images? Speak for yourself. I’m getting very good results, but that’s what you get when you put the effort in.
Masking out ghosting is quick and simple. It is easier to hide noise in detailed areas like trees and grass compared to a one tone sky which requires more images.
Horrifically time consuming? It takes me about 5 minutes from start to finish, those 5 minutes are worth it for those special shots.
I get that averaging maybe too time consuming and laborious for some, but averaging is an effective technique for those who wish to eke out every single bit of performance in image quality. My post was aimed at those people.

 

[CanadaDrone]

To echo what other people have already said, this is a very old technique that is especially difficult for drone photos that typically have more movement in them. It just doesn't work that well and isn't worth your time, especially if you want the best possible results. Even 5 minutes is far longer than it takes to get better results with other methods.

With the right software these days (like I mentioned in my earlier post), you can batch process 100 RAW photos with better results in less time than it takes to process a single image in this way. Each to their own, but if you value your time there are much better ways to go about it.