Bitrate calculations and implications for best quality

To
4: Regarding key frame distance - the DJI encoder uses a long-gap encoding technique which will generate a large number of B-frames (recording only incremental changes, not the full frame) before starting with a new I-Frame (full frame). The reason for this is mainly to allow encoding higher frame-rates (lots more data) to be fitted into the 60Mb/sec or 100Mb/sec respectively. The resulting "pulsation" can be really problematic.

I find that the flickering problem is a lot worse in mavic than in p4p. Do you know why that is? Also, why does it look like the flickering is particularly bad with trees?
 
H.265
4K:3840×2160 30 @100Mbps

H.264
4K:3840×2160 30 @100Mbps

My take on H265 vs H264 is that H265 for any identical frame rate and bitrate (such as in above example) is potentially providing up to twice the quality than H264. Naturally 60fps can only be done in H264 but for a match across both and taking into account H265 might be up to 50% more efficient than H264 for comparable quality, then essentially H265 @ 100Mbps is similar to H264 @ 200Mbps (or H264 @ 100Mbps is similar to H265 @ 50Mbps). So recording in H265 at the same fps/bitrate as H264 is going to provide up to twice the bitrate/quality to work with in post.

Is my logic correct or have I missed something?
 
My take on H265 vs H264 is that H265 for any identical frame rate and bitrate (such as in above example) is potentially providing up to twice the quality than H264. Naturally 60fps can only be done in H264 but for a match across both and taking into account H265 might be up to 50% more efficient than H264 for comparable quality, then essentially H265 @ 100Mbps is similar to H264 @ 200Mbps (or H264 @ 100Mbps is similar to H265 @ 50Mbps). So recording in H265 at the same fps/bitrate as H264 is going to provide up to twice the bitrate/quality to work with in post.

Is my logic correct or have I missed something?

Totally flawed logic. In fact you can test this yourself. You just get a smaller file in 265 not better quality.
 
Totally flawed logic. In fact you can test this yourself. You just get a smaller file in 265 not better quality.

Oh ok... so the specs are not what it will record every time just what it might record up to? So when it states @100 Mbps that's not the bit rate you will find in every file?
 
Totally flawed logic. In fact you can test this yourself. You just get a smaller file in 265 not better quality.


No that's not true -- the 4K UHD video file is exactly the same size for the same duration no matter if you use H.264 or H.265. I can't say with certainty that the video files recorded in H.265 are noticeably better but I can say with absolute certainty that they are the same size. The argument is that if H.265 is more efficient at preserving detail than H.264 AND they both are given the same bandwidth then the H.265 file should have more/better detail. So, regardless of whether you shoot in H.265 or H.264 a single 4GB file will get you 5:27 either way!


Brian
 
No that's not true -- the 4K UHD video file is exactly the same size for the same duration no matter if you use H.264 or H.265. I can't say with certainty that the video files recorded in H.265 are noticeably better but I can say with absolute certainty that they are the same size. The argument is that if H.265 is more efficient at preserving detail than H.264 AND they both are given the same bandwidth then the H.265 file should have more/better detail. So, regardless of whether you shoot in H.265 or H.264 a single 4GB file will get you 5:27 either way!


Brian

Thanks... yes that's what I thought if they are both recording at 100mbps. Visually it might not be noticeable but in theory the extra datarate is there for post work from H265. Without much motion the the scene it might not matter or have any benefit but with lots of motion/detail it possibly could depending on how DJI have worked it.
 
No that's not true -- the 4K UHD video file is exactly the same size for the same duration no matter if you use H.264 or H.265. I can't say with certainty that the video files recorded in H.265 are noticeably better but I can say with absolute certainty that they are the same size. The argument is that if H.265 is more efficient at preserving detail than H.264 AND they both are given the same bandwidth then the H.265 file should have more/better detail. So, regardless of whether you shoot in H.265 or H.264 a single 4GB file will get you 5:27 either way!


Brian

This says otherwise DJI H.264 vs H.265 - Multicopter Warehouse. This got changed recently?
 
This says otherwise DJI H.264 vs H.265 - Multicopter Warehouse. This got changed recently?

And if that article is correct then I agree. So which one is it? Has firmware updates changed the way DJI implements it. Surely someone has tested this. I mean, what is the files actual bitrate (irrelevant of file size) 100 Mbps for H264 and less for H265 or are both files 100 Mbps (and possibly slightly different file sizes). It's the bitrate that counts not file size as such. If both are 100 Mbps then H265 has the theoretical advantage.
 
Confirmed again, as per Raporman0909 above states also...

Tested today: 4K @ 30fps in H265 results in a file with bitrate of 104801 Kbps / 104.801 Mbps (that is how it is with latest firmware so all that matters right now.)

Ok but no one could show it as having a higher quality. I’ve never seen a video test to show that
 
Just to toss my $0.02 in...

I think h265 produces *slightly* better video, as in, less blur/artifacts/etc.

However, the difference is so small as to almost not be noticeable. In addition... I've noticed a higher rate of corruption/issues/etc when using h265 over h264.
In fact, it's to the point where I don't trust h265 to always produce usable files, and therefore end up using h264 most of the time.

Remember... the video engine is also doing sharpening, color profiles, and lens correction all while writing 100 mbit/s to the SD card.
h264 is clearly "easier" on the video engine (hence no 60 fps 4K w/ h265), so I also figure, why over-task it if I don't need to?

I also usually transcode h265/h264 to CineForm before grading in Resolve (I explain my typical workflows here), which seems to help a bit with the wonky stuff DJI is doing w/ their codecs.

~~~

Best advice is to try both for yourself (including at 3840x2160 & 4096x2160) and see which you think is best.

Also, FWIW, h265 transcoding is significantly slower than h264, as well as support being more limited in various NLEs and grading tools.
 
  • Like
Reactions: Dronefriend
Ok but no one could show it as having a higher quality. I’ve never seen a video test to show that


I've stated several times in the past that I can't confirm better quality using H.265 versus H.264 but the more efficient algorithm should be. DJI has not demonstrated to me that they are competent with image processing and the continued fumbling with D-log is all the evidence I need on that score. I have also stated numerous times that video compression techniques is not for the novice technician and is in fact among the most complex things around. You don't get to work on the guts of it without a very high level PHD in math -- there is some daunting science/math behind image/video compression. So, few of us have what it takes to understand let alone pioneer in the field and sadly, when many tech websites talk about it they do so with all the detail of a Readers Digest summary. The approach taken reminds me of the Monty Python skit "How to do it" -- in one such skit they explain how to play the Flute, "Blow in the end and move your fingers over the holes". Accurate at one level but completely useless.

I think the idea that H.265 produces smaller files is from just such a cursory understanding of what H.265 is. A Google search tells us that it needs less data to code the same video with the same quality. The website that was mentioned did just that and apparently didn't think to actually measure the file sizes.


Brian
 
I've stated several times in the past that I can't confirm better quality using H.265 versus H.264 but the more efficient algorithm should be. DJI has not demonstrated to me that they are competent with image processing and the continued fumbling with D-log is all the evidence I need on that score. I have also stated numerous times that video compression techniques is not for the novice technician and is in fact among the most complex things around. You don't get to work on the guts of it without a very high level PHD in math -- there is some daunting science/math behind image/video compression. So, few of us have what it takes to understand let alone pioneer in the field and sadly, when many tech websites talk about it they do so with all the detail of a Readers Digest summary. The approach taken reminds me of the Monty Python skit "How to do it" -- in one such skit they explain how to play the Flute, "Blow in the end and move your fingers over the holes". Accurate at one level but completely useless.

I think the idea that H.265 produces smaller files is from just such a cursory understanding of what H.265 is. A Google search tells us that it needs less data to code the same video with the same quality. The website that was mentioned did just that and apparently didn't think to actually measure the file sizes.


Brian

So much this... there are TONS of "reviews" of the P4P talking about the "smaller file size" the h265 codec enables.
Complete BS, as any P4P owner and attest to. It's like the reviewers just Googled "h265 vs h264" and put down a sentence summarizing the first couple search results.

~~~

To be honest, I think the current P4P implementation of the whole Ambarella H1 is somewhat borked.
They were one of the first in the market to use it... and as you say, DJI does not have an excellent track record of implementing effective video encoding.
 
Just to toss my $0.02 in...

I think h265 produces *slightly* better video, as in, less blur/artifacts/etc.

However, the difference is so small as to almost not be noticeable. In addition... I've noticed a higher rate of corruption/issues/etc when using h265 over h264.
In fact, it's to the point where I don't trust h265 to always produce usable files, and therefore end up using h264 most of the time.

Remember... the video engine is also doing sharpening, color profiles, and lens correction all while writing 100 mbit/s to the SD card.
h264 is clearly "easier" on the video engine (hence no 60 fps 4K w/ h265), so I also figure, why over-task it if I don't need to?

I also usually transcode h265/h264 to CineForm before grading in Resolve (I explain my typical workflows here), which seems to help a bit with the wonky stuff DJI is doing w/ their codecs.

~~~

Best advice is to try both for yourself (including at 3840x2160 & 4096x2160) and see which you think is best.

Also, FWIW, h265 transcoding is significantly slower than h264, as well as support being more limited in various NLEs and grading tools.

Thanks for this info... if it turns out less reliable recordings in critical times then agree any possible compression/bitrate advantage is not worth the risk. And yes I will do exactly that and test/compare if H265 does actually afford any worthwhile difference @ the same size/framerate. Naturally it always helps to see if others have already done such and found a definitive answer.
 
My findings...

Drone.jpg
 
Great table.

However, if you don't mind my asking... What's your subjective takeaway after testing all this?

For now 4K 30fps H264 is my go to mode. Still want to test 2 things properly to make a final opinion:

1. Fast movement and detailed scenes (eg flying low over long grass). It has been reported elsewhere that 1920x1080 and 2.7K retain much more detail is such scenes and movement than 4K. I trust this is the case but want to test for myself.

2. Does H265 have any benefit over H264 or visa versa. Opinion is H264 is the better option.
 
  • Like
Reactions: skyeboysteve
For now 4K 30fps H264 is my go to mode. Still want to test 2 things properly to make a final opinion:

1. Fast movement and detailed scenes (eg flying low over long grass). It has been reported elsewhere that 1920x1080 and 2.7K retain much more detail is such scenes and movement than 4K. I trust this is the case but want to test for myself.

2. Does H265 have any benefit over H264 or visa versa. Opinion is H264 is the better option.

Thanks for the insight! Look forward to your final conclusions!

And thanks for sharing with the community!
 

Members online

No members online now.

Forum statistics

Threads
143,066
Messages
1,467,354
Members
104,933
Latest member
mactechnic