Extreme pixelation/aberration | P4P

[Kristina Fowler]

Note extreme pixelation. It's really bad.

Source file: 2720x1530; 30FPS; 2-Pass; Target Bitrate 55; Max Bitrate 60; D-log + ND8 filter

Rendered file: 1920x1080

Post: I did apply a LUTS + shadow removal + sharp; modest amount of sharp, contrast & vibrance.

The only thing I can think of is write speed of micro SD card (using SanDisk Extreme Pro, rated at 90MB/s write).

Any other suggestions? T.I.A.

KF

 

[Kristina Fowler]

Files are source files of the microSD Card straight out the Phantom 4 Pro (P4P). Vids are not recorded on iPad Pro.

 

[PaulSouthport]

What's the raw footage like?

 

[Kristina Fowler]

What's the raw footage like?

Raw footage looks fine. I understand when you upload to Youtube/Vimeo, it strips some color, but this is pretty bad stuff.

 

[PaulSouthport]

OK, and if you play the rendered file on your PC?

 

[eaglegoaltender]

Any other suggestions? T.I.A.

Hi KF........I was having same issues when trying out 4k and sometimes at 1080p.

I found some very good info relative to this here Video grading pixelation issue

Good luck and safe flying - very good job on the vid while flying!

 

[LarBear360]

To me it looks like an issue with the processed file. Too much detail removed to create a YouTube Friendly file?

I didn't see a reply as to how the processed footage looks on your PC.

Go back to the original un-retouched video file and try uploading that.

 

[Raptorman0909]

Yeah, that's not what I'd call "pixelation" but more like "blocking up" due to compression.

Also, I have had only negative experience using D-log and have pretty much settled on using D-Cinelike with the following settings: -1, 0, -1 or -2, 0, -2 for custom settings, I stay at ISO100 whenever possible and generally use f/5.6. I do not use ND filters and don't feel a need to do so and my shutter speed, controlled manually, will run from 1/320 to 1/2500 depending on lighting with most daylight situations averaging around 1/1000. I do not have issues with rolling shutter that many think you'll get with higher shutter speeds. I shoot in 4K30 (UHD) -- when I first started with drones I was shooting in 4K24 but the lower frame rate lead to unpleasant stuttering and going to 30fps made a big difference. Going faster than 30fps at this stage is not recommended due to bandwidth and processor limitations in the drone.

In post I'd recommend beginning without using LUT's and avoid the tendency to exaggerate contrast or sharpness. Over time you can add LUT's but get a handle on editing without them first.

As an example of what the P4P can achieve with the settings I mentioned check this out.

Brian

 

[Kristina Fowler]

Yeah, that's not what I'd call "pixelation" but more like "blocking up" due to compression.

Also, I have had only negative experience using D-log and have pretty much settled on using D-Cinelike with the following settings: -1, 0, -1 or -2, 0, -2 for custom settings, I stay at ISO100 whenever possible and generally use f/5.6. I do not use ND filters and don't feel a need to do so and my shutter speed, controlled manually, will run from 1/320 to 1/2500 depending on lighting with most daylight situations averaging around 1/1000. I do not have issues with rolling shutter that many think you'll get with higher shutter speeds. I shoot in 4K30 (UHD) -- when I first started with drones I was shooting in 4K24 but the lower frame rate lead to unpleasant stuttering and going to 30fps made a big difference. Going faster than 30fps at this stage is not recommended due to bandwidth and processor limitations in the drone.

In post I'd recommend beginning without using LUT's and avoid the tendency to exaggerate contrast or sharpness. Over time you can add LUT's but get a handle on editing without them first.

As an example of what the P4P can achieve with the settings I mentioned check this out.

Brian

Brian - thanks for the tip. I've since edited a few clips (not yet uploaded) with nothing more than tweaking the curve. It actually looks a bit more realistic. Perhaps I was guilty of slopping on the LUTS. I'll also try the original LUTS, but with saturation turned down and do an A/B comparison. My thinking on D-log is that it's the purest capture possible since the footage bypasses any "treatment" that the P4P camera applies to it. By doing so (theoretically), it gives me total control over the final output by applying my own color correction. I'm sure at some point I'll throw in the towel, but I want to experiment a while in this direction.

 

[Kristina Fowler]

Please see the video below. Using PremierePro with same export settings as before.The left side is the original video (with LUTS). The right side is no LUTS, lightly tweaked the curves, other modest adjustments (Exposure +1, Contrast +10, Highlights +10 [clouds, Sharpness +5). I resisted the temptation to remove any shadows. I consider these adjustments to be benign. My main concern is the grassy areas - where there is very subtle color shifts. Muted browns and greens. I really don't think there is any measurable difference in the pixelation issue from the left side to the right side. It's still pretty bad. I think the camera simply has issues discerning between the very gradual color shifts, which I guess is another way of saying dynamic range issues. It also seems to be more pronounced as you get closer to the edge of the frame. Maybe it's the Neutral Density Filter. Very frustrating!

 

[LarBear360]

You said earlier your raw footage looks fine. I think your problem is with YouTube. Try this and see what happens:

In Publish+Share, select Computer, then AVCHD.

In Presets, select MP4 - H264 1920x1080p 30 (or 24 or 25, depending on your project/source fps).

Then click Advanced.

In the Video tab:

• Change Profile to High.
• Check Render at Maximum Depth.
• Change Bitrate Encoding to VBR, 2 Pass (although 1 Pass is usually almost as good).
• Change Target Bitrate (Mbps) to 4.
• Change Maximum Bitrate (Mbps) to 16.

 

[Kristina Fowler]

LarBear360 - not really understanding where your heading with this. Do you mean the settings in PremierePro? Can't find any of these inside YouTube settings....

My PremierePro settings align with these except for the Bitrate. I always thought more is better for bitrate. Mine are Target 55/Max 60.

Are we both talking about PremierePro???

 

[LarBear360]

LarBear360 - not really understanding where your heading with this. Do you mean the settings in PremierePro? Can't find any of these inside YouTube settings....

My PremierePro settings align with these except for the Bitrate. I always thought more is better for bitrate. Mine are Target 55/Max 60

Yes, try lower settings in Premier and see what happens. LUTs, color correction, etc. should not do that. I think something is happening with YouTube.

 

[Kristina Fowler]

I tried the default PremierePro settings for "YouTube 1080p HD" --->>> 1 Pass; bitrate 16/16; render at max depth = unchecked. Here's what I got:

It's especially bad in the upper 20% of the frame.

Settings:

 

[LarBear360]

Hmmm. Not my idea but try changing the following I found from a custom profile on another site

Under “Bitrate Settings” set “Bitrate Encoding” to “CBR” and your “Target Bitrate [Mbps] to “40.” Lastly, check “Use Maximum Render Quality” at the bottom.

If you really want to drill down, send me 60 seconds of your raw un-touched file just as it came off the card. I'll test in Vegas, Cyberlink, and Adobe Premiere Elements which is the consumer-friendly version of Pro.

You would have to upload to a drop-box or something and post a link or I can message you an e-mail you can use.

 

[Raptorman0909]

Brian - thanks for the tip. I've since edited a few clips (not yet uploaded) with nothing more than tweaking the curve. It actually looks a bit more realistic. Perhaps I was guilty of slopping on the LUTS. I'll also try the original LUTS, but with saturation turned down and do an A/B comparison. My thinking on D-log is that it's the purest capture possible since the footage bypasses any "treatment" that the P4P camera applies to it. By doing so (theoretically), it gives me total control over the final output by applying my own color correction. I'm sure at some point I'll throw in the towel, but I want to experiment a while in this direction.

No D-log actually does a great deal of modification to the video -- it dedicates most of the levels to mid-tones leaving fewer levels for highlights and shadows. D-Cinelike actually does less modification to the video. D-log flattens the video and that IS a modification.

I would shoot the video in 4K30 and when you do that the drone records at 100Mbps -- the best it can do. When rendering, and it looks like you're using Adobe PP, I'd try rendering as 4K30 and use a bit rate of about 60Mbps with the YouTube 2160P 4K setting. Then upload to YouTube and then view the video in whatever resolution you can with your monitor -- preferably 4K but I'm not sure what your monitor is capable of. Be aware that it can take as much as a day or even more for a 4K video to be available in 4K so at first it may only be available as 480 or less.

The blocking up I was talking about is a sure sign of excessive compression. Prior to the P4P I had the Inspire I Pro drone with X5 camera and when using the best settings I could get at 4K30 things like foliage would tend to block up as is common with low bit rate dashcams. Moving to the P4P and using the same settings there is MUCH less blocking up and that is down to the higher bit rate of the P4P (100Mbps versus 60Mbps).

In your case I suspect the rendered video was rendered using too low of a bit rate.


Brian

 

[msteele]

Good info here for a novice such as myself. I'm also curious as to how the 60 sec video from LarBear360 turns out if Kristina sent it to him.

 

[LarBear360]

Good info here for a novice such as myself. I'm also curious as to how the 60 sec video from LarBear360 turns out if Kristina sent it to him.

She has not, however, the offer still stands. I'm curious myself as to what's causing this.