H.265 VS H.264 - D-Cinelike comparison

Joined
Oct 7, 2013
Messages
114
Reaction score
70
Location
Piraeus, Greece
Comparison footage in 4K between H.265 and H.264 codecs with D-Cinelike in DJI Phantom 4 Pro.
ISO 100, Style: Sharpness -1, Contrast -3, Saturation -2

I think we must use h.265 in D-Cinelike because we get more accurate colors, less artifacts and less noise than in h.264

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


Kostas, nice work, confirms work I did a while back with H.265 and D-log even though the version of D-log I used back in January was not the same.

YouTube

The banding issue is very evident with H.265 and less so with H.264 making H.264 preferred when using D-log, however, the sky region using D-log and H.264 shows much more noise than D-log with H.265.

As for me I've pretty much given up using D-log at all and have settled on D-Cinelike and been reasonably happy using it. Of course, banding is always going to be an issue with an 8-bit camera and D-log just makes that problem worse by allocating more of the levels for mid-tones leaving fewer for highlights.


Brian
 
Kostas, nice work, confirms work I did a while back with H.265 and D-log even though the version of D-log I used back in January was not the same.

YouTube

The banding issue is very evident with H.265 and less so with H.264 making H.264 preferred when using D-log, however, the sky region using D-log and H.264 shows much more noise than D-log with H.265.

As for me I've pretty much given up using D-log at all and have settled on D-Cinelike and been reasonably happy using it. Of course, banding is always going to be an issue with an 8-bit camera and D-log just makes that problem worse by allocating more of the levels for mid-tones leaving fewer for highlights.


Brian
You are absolutely right!
 
Comparison footage in 4K between H.265 and H.264 codecs with D-Cinelike in DJI Phantom 4 Pro.
ISO 100, Style: Sharpness -1, Contrast -3, Saturation -2

I think we must use h.265 in D-Cinelike because we get more accurate colors, less artifacts and less noise than in h.264

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

I am not seeing the improvements in colors, artifacts and noise in your videos at all. They look identical, apart from the trees looking sharper in left upper corner in h264 and the overall picture less detailed in h265.

I am very interested in the subject, have you conducted any other tests? Thanks!
 
The intent of H.265 is to reduce by a third the file size compared to H.264, not to improve image quality. The purpose is to require less bandwidth for 4k video streaming. The downside of H.265 is that it takes 10 times the processing power to decompress the file which is why it is not likely to be a mainstream format for some time to come.

How DJI engineers and others have implemented H.265 and H.264 in their video encoding is another matter entirely. The output file to a standard format is not the same as having all applications doing the data compression in the same manner. There are also going to be differences in how a computer application processes the two formats and how well a GPU processes the video and how it is displayed on the computer device's display.

Even with a given vendor's compression software there are usually multiple options for the compression with one trading speed over quality for faster data compression rates. I would expect a different compression routine to be used at 60fps for example than at 30fps to compensate for the drone CPU having to process data twice as fast as it is doing this in real time.

H.265 may do a better job of masking noise from the original data but if that is the case then it is to be expected that it is also reducing detail and possibly dynamic range as well. I can see the value of doing some testing as the OP has done to see with ones own eyes if there is a discernible difference and in what situations. All it costs is a little bit of time, but this test would need to be redone with each new DJI AC firmware release to verify nothing has changed in this area.
 
All bets are off as far as DJI's implementation of, well, anything, and they have not displayed great competence at image processing generally. If DJI implemented H.264 and H.265 in an appropriate manor AND both streams are stored at 100Mbps then there SHOULD be more detail and/or fewer obvious compression artifacts using H.265 over H.264. Assuming both are properly done then an equivalent video file in H.265 should require less than the 100Mbps that H.264 uses to achieve the same IQ. But again, all bets are off as to how well DJI coded them.


Brian
 
The intent of H.265 is to reduce by a third the file size compared to H.264, not to improve image quality.

I don't think that's correct, or at least not in this particular case. H.265 can be used to reduce file size for a similar quality, or it can be used to improve quality within the same bandwidth. In the case of the P4P resultant files sizes are similar so the intent is clearly not smaller file sizes, rather since the maximum bandwidth is limited to 100 mbps then H.265 should exhibit better quality within this limitation than H.264. Whether it actually does or not (due to DJI's implementation, good or bad) may be questionable, but the intent in this particular application is higher quality.
 
Last edited:
To the point, which is better overall, in real use? i understand it takes a lot more processing power to work with 265.
 
A year down the line here, I wondered if folks had more thoughts and experiences on the comparison between the two codecs and reached a decision as to which is preferable ? I need to cut in a selection of P4P aerial material in a 4k docie project where the principal cameras have been Red Dragons, a tough challenge I appreciate ? I have done some experimenting on both codecs using D-Cinelike rather than D-log. What worries me a little is having to use ND16 most the time when shooting on 265 25p and 50' shutter with trying to keep an aperture setting close to F5.6 rather than say F11. Any thoughts from the quality gurus and perfectionists most welcome ? :)
 
A year down the line here, I wondered if folks had more thoughts and experiences on the comparison between the two codecs and reached a decision as to which is preferable ? I need to cut in a selection of P4P aerial material in a 4k docie project where the principal cameras have been Red Dragons, a tough challenge I appreciate ? I have done some experimenting on both codecs using D-Cinelike rather than D-log. What worries me a little is having to use ND16 most the time when shooting on 265 25p and 50' shutter with trying to keep an aperture setting close to F5.6 rather than say F11. Any thoughts from the quality gurus and perfectionists most welcome ? :)
I know it doesn't answer your question about codecs but absolutely use the ND to keep your aperture lower. As far as I'm concerned, keeping the f-stop in the lens sweet spot is a hard fast rule for professional looking drone video. That's especially true if you're going to cut it with a digital cinema camera. Personally I haven't seen a visible difference in my tests between h.264 & h.265. That's just been my experience.
 
  • Like
Reactions: foxtrot
Yeah agree, but all the same it seem nuts having to darken the image so much, kind of reminds me about the film days and before daylight balanced stock with having to shoot with a Wr 85 filter !
No doubt technology will improve and change for the better and we'll then reminisce about the days we used to pack NDs on front of our phantom lenses, roll on Phantom 6 pro or will we have to wait until P10 P....:)
 
Comparison footage in 4K between H.265 and H.264 codecs with D-Cinelike in DJI Phantom 4 Pro.
ISO 100, Style: Sharpness -1, Contrast -3, Saturation -2

I think we must use h.265 in D-Cinelike because we get more accurate colors, less artifacts and less noise than in h.264

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

I doubt that Adobe Premiere CS6 can even import h.265. Who can even edit that? My dual Xeon E5-2667 with 128GB RAM struggles with h.264 4K footage at 24 FPS as it is. I can't imagine trying to edit h.265.
 
H.265 requires 10x the processing power to compress and decompress the data and that is why it is not likely to be widely used for many years. A smaller file is of less value when the decompression adds more time than is saved by transmitting a smaller file.

The intent of H.265 was to create a smaller file with minimal impact on image quality with the advent of 4K and higher resolution recording taking place. 4K creates 4x as large a file as 1080p for the same recording time and there are recorders now providing 5K and soon 8K recording and so there was and is a need to compress the data as much as possible so as not to overwhelm the data drives in use.

With Apple 4K ProRes 4444 XQ a production camera shooting 100 hours of footage can produce 76 TB of data and all that data needs to be stored somewhere. The Phantom Flex 4K RAW at 24fps produces 1500 GB/hour and numbers like this are what is driving the need for data compression that is effective and efficient.

With the relatively low amount of data coming off a consumer level drone camera there is no need to be using H.265 for better recording data compression.
 
Good information. I work in Prores most of the time and we have terabytes of storage for media when working with masters from 35mm cinema cameras. In a way, I wish the Phantom shot Prores. It's the most compatible format. It is one of the factors driving me toward getting an Inspire 3 when it hits the market.
 
  • Like
Reactions: Ripper7620

Members online

No members online now.

Forum statistics

Threads
143,066
Messages
1,467,358
Members
104,936
Latest member
hirehackers