Is 4k60 video different from 1/2.3" CMOS vs 1"CMOS?

[Barton]

You doubtless know this but the EVO app is optimised fr either iOS 11.4 or Android 8 or later.

Not sure what 'optimized' means for this app. The Google Play Store says it it will run on Android 4.4 and later. My refurb phone is at Android 6 and my primary smartphone is at Android 7. Android 8 (Oreo) was released a year ago (Aug 21, 2017). Here in the USA, it can take over a year for phone providers to upgrade their phones to the latest Android OS. So far, the Autel Explorer app seems to run fine...

 

[YuKay]

Not sure what 'optimized' means for this app. The Google Play Store says it it will run on Android 4.4 and later. My refurb phone is at Android 6 and my primary smartphone is at Android 7. Android 8 (Oreo) was released a year ago (Aug 21, 2017). Here in the USA, it can take over a year for phone providers to upgrade their phones to the latest Android OS. So far, the Autel Explorer app seems to run fine...

From the horse's mouth:

Hello YuKay,

The Autel Explorer app is optimized for iOS 11.4 and Android 8 so anything older than that may not function at its best capacity. Not to say it will not work but I would recommend newer devices if they are within your budget.

Best,

Alejandro R.
Customer Support
AUTEL ROBOTICS

 

[Barton]

The JPEG is over-sharpened but nonetheless looks a lot better than the (surely faulty) H Plus in-camera JPEG.

At 5.7MB, the EVO 12 megapixel JPEG can't be directly compared to oliver's previously uploaded 15.0MB 20 megapixel H Plus JPEG but I compared them anyway. I see more colour and less noise in the EVO image and significantly fewer sharpening artifacts. Compare the ugly white halos in the two 800% zooms below where the dark landscape meets the light sky. It concerns me that Yuneec haven't fixed their obvious JPEG algorithm blunder yet.

View attachment 11450

It also appears that Autel have graded their sharpening algorithm so that the artifacts are less noticeable in the foreground, resulting in a much more usable image.

I also compared the poster frame from your EVO 4k/60 video with that of oliver's H Plus YUN_0003.MP4 4k/60 video. Both are compressed with the H.264 codec but for me, the sharpening halo is more pronounced in the H Plus image below:

View attachment 11451

The halo around the foreground balcony post in the H Plus screen grab above may be an extreme example but for my money, EVO also grades the foreground to distance sharpening algorithm in its video too, making them easier and more natural to look at. The (MP4 screen grab at 800%) example below shows EVO's sharpening halo applied at different levels to the foreground foliage and the distant tree:

View attachment 11453

That said, many people won't see the MP4 halos from either camera at 100% view unless they have a very sharp eye and are looking for them.

As things stand, for me, the EVO output beats the H Plus in some key departments, while costing less; and, for some reason, its 4k/60 video plays a lot more smoothly on Youtube. If I didn't need the more pro-oriented toolset of the H Plus, its truly standalone controller, its bigger still images and its superior in-camera colours - and its official UK availability - it would be a no-brainer!

@YuKay Very interesting. I'm now starting to understand the level of detail analysis that you are applying to the images .

 

[Barton]

Barton, I have enhanced/ruined a still from your video. No pixels were sharpened in this 33-second Photoshop dash - which speaks for the clarity of the EVO footage…

View attachment 11459

@YuKay Very dramatic colors - certainly a huge change from the original. Time to up my post processing game . Thanks for showing me what is possible.

 

[YuKay]

@YuKay Very dramatic colors - certainly a huge change from the original. Time to up my post processing game . Thanks for showing me what is possible.

It's the Dutch School of Taking Liberties with Light and Shade .

If I want more of either, I just add a Photoshop spotlight (a bleedin' liberty, as we Brits might say). The rest is a cocktail of Curves, Saturation, Vibrance, Exposure, Colour Balance, Channel Mixer and Selective Colour.

 

[YuKay]

I have now forensically compared Barton's EVO 4k/60 footage with the H Plus equivalent can can find no real differences and certainly no possible explanation for the former playing smoothly for me on Youtube while the latter does not.

For now, I will have to assume that the EVO's Ambarella H2 video processing chip is more efficient than that in the H Plus. Actually, I think the C23 may have a version of the H2 chip too but maybe not the same version?

Otherwise, I'm stumped.

 

[rdonson]

We’re both h264 or h265?

 

[YuKay]

We are Ron. And the E90 uses the Ambarella H2 so, logically, so does the C23…or a version of it.

I'm clutching at straws.

Welcome back! Safe and dry I hope…

 

[YuKay]

This is a new H Plus 4k/60 first flight video uploaded by Shawn Ide Studios and there's much to like, especially the colours. But for me, it is less sharp than Barton's EVO 4k/60 download (which hasn't been degraded by Youtube of course) and it contains many more antialiasing artifacts, especially in the always-challenging tree foliage. Also the gimbal pans play here with an unacceptable stutter. I would like to see a clip of the source footage for a proper comparison to be made.

(EDIT: I see that Shawn is going to upload a source footage clip. The stuttering does improve at 1080p60 view - or maybe it's helped by the cache - but is still visible even at 480.)

It did drop fewer frames (215 dropped out of 52,000) than most H Plus 4k/60 videos on my rig but the smoother playback was offset by the unsmooth pans. I also think the video is over-sharpened. The halo along the top the distant mountains at 2:25 is clearly visible at 100% view and it should not be.

 

[AveryBadman96]

If the stuttering improves when you watch it at 1080p I suspect the problem is what you are viewing it on. If it is a computer or tv, does your computer graphic card support 4k+ resolution, or are you maxing out your RAM running programs at the same time as viewing the video? Or is your TV or computer monitor able to display 4k video? Most 4k tv's 4k is 3840x2160 not 4096x2160. I know whenever i had stuttering on my videos i looked at 3 things: my monitor capabilities, my computer graphic card, and/or SD card write speed. I record in 4096x2160, i have to down sample to 2k for my tv, however my computer plays beautifully because my graphics card is capable of viewing at 7680x4320, and I have 64gb RAM and monitor views at 5k. No lag or stutter.

 

[YuKay]

If the stuttering improves when you watch it at 1080p I suspect the problem is what you are viewing it on. If it is a computer or tv, does your computer graphic card support 4k+ resolution, or are you maxing out your RAM running programs at the same time as viewing the video? Or is your TV or computer monitor able to display 4k video? Most 4k tv's 4k is 3840x2160 not 4096x2160. I know whenever i had stuttering on my videos i looked at 3 things: my monitor capabilities, my computer graphic card, and/or SD card write speed. I record in 4096x2160, i have to down sample to 2k for my tv, however my computer plays beautifully because my graphics card is capable of viewing at 7680x4320, and I have 64gb RAM and monitor views at 5k. No lag or stutter.

My rig is capable - it's a 5k 27in retina iMac which can play video natively at greater than 4k. I have 40GB of RAM and 4TB of internal SSD disk so it's probably capable of doing a lot more than I have yet been able to throw at it.

The more likely culprit wrt Youtube playback is my internet connection which has recently been beefed up to around 45-50Mbps although it does vary. And I join my network via wi-fi which lowers the speed a bit. Even so, I get 25-40Mbps all the time.

In any event, Youtube 4k30 video from any source invariably plays smoothly. What causes stuttering is H Plus 4k video at 60fps. This might indicate that my set-up isn't capable of playing Youtube 4k60 video smoothly…and yet it does always play smoothly 4k60 video which has been shot with the Autel EVO.

And that, in a nutshell, is the mystery.

 

[YuKay]

I have been looking further afield to try to understand why a 1/2.3" CMOS (such as the Autel EVO) might (imho) produce clearer 4k video with fewer artifacts than a 1"CMOS (such as the H Plus).

In this video from Captain Drone, the 4k output from the EVO (1/2.3") is compared with that from the Mavic Pro 2 (1") and while there are some funny things going on with over-saturation in the Mavic footage (and it is definitely better in low light), it is clearly less sharp than the EVO footage and it suffers more from antialiasing in the grass and foliage. (The sharpness difference is pretty obvious in the row of houses from about 5:50 and it's not just a depth of field thing as foreground objects are less sharp in the Mavic footage too.)

So now I'm starting to wonder if this is a wider issue and that some current 1" sensors have to rely on hacks and compromises to produce 4k video. Certainly, some folk on the MavicPilots forum think so; and Cliff Totten has made an interesting analysis video which seems to show that the real resolution of the MP2's standard 4k video is closer to 2.7k - which he suspects is due to sensor overheating requiring line skipping or pixel binning. Whether the H Plus sensor and processor make similar compromises, I don't know…but I do think the smaller sensor of the EVO produces clearer and cleaner 4k video than the 1" sensor of the H Plus.

 

[Ty Pilot]

Interesting @YuKay I just viewed CD's video comparison (using that word real loosely here) and if that were my MP2 it would go back to whence it came ASAP. Seriously though, I think beyond what is happening at the back end of capture process, (sensor) I think whats at the beginning of them may also be playing a role. In fact we've seen what a Sony 1 inch sensor can do behind professional glass and I think it is the lens that is the weak spot in aerial photography. It seems the Hasselblad may need some tweaking, as I don't think you could sharpen the MP2 footage in post to match the EVO at least not from CD's footage.

I do find it interesting that you think the Evo produces cleaner footage at 4K than the H Plus. I've not been following the Evo since I have had too much Plus Koolaid but do you have an example of Evo footage, in particular; that you find shows advantages over the Plus?

 

[YuKay]

Interesting @YuKay I just viewed CD's video comparison (using that word real loosely here) and if that were my MP2 it would go back to whence it came ASAP. Seriously though, I think beyond what is happening at the back end of capture process, (sensor) I think whats at the beginning of them may also be playing a role. In fact we've seen what a Sony 1 inch sensor can do behind professional glass and I think it is the lens that is the weak spot in aerial photography. It seems the Hasselblad may need some tweaking, as I don't think you could sharpen the MP2 footage in post to match the EVO at least not from CD's footage.

I do find it interesting that you think the Evo produces cleaner footage at 4K than the H Plus. I've not been following the Evo since I have had too much Plus Koolaid but do you have an example of Evo footage, in particular; that you find shows advantages over the Plus?

Not a like-for-like comparison, Ty Pilot but I could link you to a number of EVO Youtube videos which play smoother, sharper and cleaner on my screen than H Plus comparables. Best though to look at source footage and there isn't much on offer. Earlier in this thread Barton kindly uploaded some EVO source footage from his maiden flight and, for me, it confirms my Youtube suspicions in that there is less noise, less sharpening, fewer artifacts and less antialiasing in Barton's video than in any H Plus source footage I have seen to date.

Yesterday, Shawn Ide uploaded some short H+ source clips of similar subject matter from his maiden flight and I found the antialiasing in them very distracting - and overall the EVO footage seemed smoother and clearer. Since both craft were new I assume they were both using latest firmware (which makes a big difference to the EVO as any online footage before September shows a stuttering gimbal issue as well as a tilted horizon).

Highly detailed tree foliage, grass, gravel and the like are difficult to capture cleanly and probably no prosumer drone camera makes a perfect job of it. But, for me, the EVO does it better than any other sub-$2k drone that I have looked at (…and significantly better than any other sub-$1k drone).

From still frames of both that I have examined at 800% view, it is clear to me that the EVO suffers less from capture, compression and sharpening artifacts than the others (as per my previous illustrated post).

 

[Rubik]

This is a new H Plus 4k/60 first flight video uploaded by Shawn Ide Studios and there's much to like, especially the colours.
It did drop fewer frames (215 dropped out of 52,000) than most H Plus 4k/60 videos on my rig but the smoother playback was offset by the unsmooth pans. I also think the video is over-sharpened. The halo along the top the distant mountains at 2:25 is clearly visible at 100% view and it should not be.

YuKay, here's my impressions from Shawn's video:
I had to do a screen grab from the video and zoom in to see the "halo" on the mountains at 2:25. Yes, it's there. Some sharpening was apparent in the foreground as well. But, as the video was playing on my 4k monitor, the sharpening was not noticeable to me. Having pictures stream by at 60/sec covers a lot of ills.
The only stuttering I saw (very rare) was remedied by backing up a bit and replaying the sequence. That implies the YT streaming or my internet connection couldn't quite keep up 100% of the time.
Update: I looked at the 4k60 short clip from Shawn. Again, I zoomed in on a 4k screen grab to see the sharpening halos. The antialiasing seems to be at the single pixel level and, therefore, unavoidable.
I agree that the EVO footage I've seen looks great and the M2P footage looks inferior. I guess we need that direct comparison between TH+ and M2P and EVO in appropriate modes and settings.

 

[YuKay]

YuKay, here's my impressions from Shawn's video:
I had to do a screen grab from the video and zoom in to see the "halo" on the mountains at 2:25. Yes, it's there. Some sharpening was apparent in the foreground as well. But, as the video was playing on my 4k monitor, the sharpening was not noticeable to me. Having pictures stream by at 60/sec covers a lot of ills.
The only stuttering I saw (very rare) was remedied by backing up a bit and replaying the sequence. That implies the YT streaming or my internet connection couldn't quite keep up 100% of the time.
Update: I looked at the 4k60 short clip from Shawn. Again, I zoomed in on a 4k screen grab to see the sharpening halos. The antialiasing seems to be at the single pixel level and, therefore, unavoidable.
I agree that the EVO footage I've seen looks great and the M2P footage looks inferior. I guess we need that direct comparison between TH+ and M2P and EVO in appropriate modes and settings.

Thanks Rubik. Maybe it's my retina 5k screen but I can see the sharpening halo at 100% on Shawn's source footage. But as I opined earlier, most probably wouldn't notice it. I can also see it on Barton's source footage but it is less pronounced.

I believe that both Shawn and Barton's source footage was shot at default settings with no post processing so the comparison is a fair one. I'm not interested enough in the MP2 to go hunting for source clips - and even less interested after seeing Captain Drone's MP2-vs-EVO comparison.

The stuttering in Shawn's pans was also evident in the downloaded source footage, albeit only at a couple of points. The EVO horizontal pans appeared to be more fluid although the up/down tilt isn't perfect yet.

The antialiasing in Shawn's H Plus source video, especially in the trees and grass, I found much more distracting than in Barton's EVO footage. But yes, all these videos are shot under different conditions and probably at different drone speeds so a direct comparison isn't possible. However, I have downloaded a good number of H Plus 4k60 source clips from different places and times and I have been distracted by the antialiasing in all of them. Tbh, I wasn't overly impressed by the EVO footage I saw until Autel's latest firmware update fixed most of the problems. So maybe Yuneec can do something similar.

But for now, I prefer the 4k60 video from the EVO.

 

[PatR]

So how much of this might be attributed to differences between the h264 and h265 codecs? I have read that only newer computers will be able to deal with h265 and many processing programs are not capable of working with it right now.

 

[YuKay]

So how much of this might be attributed to differences between the h264 and h265 codecs? I have read that only newer computers will be able to deal with h265 and many processing programs are not capable of working with it right now.

PatR, I haven't found very much H.265 H Plus source footage so my observations are based on H.264. There are a few useful C23 comparison clips here but I don't see much difference in quality or playability between the H.264 and H.265 files except that the latter benefit from a slightly warmer colour palette.

For the avoidance of doubt, I have seen some very high quality H Plus 4k footage - just not at 60fps. And in a flat light with ultra-slow camera movements I have also seen a very clean 4k/50fps video (German village) - albeit with stuttering pans.

For whatever reason, the H Plus seems to struggle with higher frame rates while the EVO handles them better, imho. The EVO system also seems to cope better with camera movement which is faster than dead slow.

Incidentally, I looked at the 1080p and 720p 120fps examples at the link above and found them unwatchable. Does anyone else on a different computer think they are of acceptable quality?

 

[YuKay]

Another enhanced/ruined still from Barton's EVO footage…

 

[PatR]

There are times I think we want more from these small systems than they can possibly deliver for the price paid. A friend of mine makes rigs that provide some of the best imagery money and user talent can deliver but that comes with a high price tag, before adding for the camera and gimbal of choice. Unless I have a ready customer base that will pay a heck of a lot more for the effort and product it’s impossible to justify going off the deep end for this stuff. I have to remember most customers satisfy a lot more easily than with what it takes to satisfy me.