JPEG/MJPEG vs HEVC

[SMGJohn]

Someone wanted to see what JPEG 15 fps image burst was compared to the HEVC, and to say the least the difference is massive not just in the dynamic range but the pure colour information presence in the images, now I have chosen to use 1080p to show the difference as much as possible because its not as noticeable in 2160p, the only difference at such high resolution is the dynamic range, now its clear that JPEG will always be superior to HEVC in 1080p mode and perhaps even in high resolution as well simply because they contain more information per frame. The JPEG have superior resolution as well compared to HEVC even in 1080p, both JPEG and HEVC have the same settings applied to them including sharpness.

Both HEVC and JPEG where colour graded in After Effects using the GammaDR2LOG -> VisionLog - LOG Generic_FujiColor 200_CIN -> Cineon to Film Contrast 01 -> Lumetri Adjustments

In camera settings used for both HEVC and JPEGs are: 1/100 - F/4 - ISO100

JPEG are super fine and resolution: 2048x1152 (Downsampled to 1080p in After Effects)

HEVC are Pro quality and resolution: 1920x1080

 

GammaDR settings:

Red X1.00 - Green X0.95 - Blue X1.00

Sharpness -10

Contrast -10

Luminance 16-235

Master Pedestal not touched

Picture Style Normal settings:

Red X1.00 - Green X0.95 - Blue X1.00

Sharpness -7

Contrast -10

 

HEVC LEFT - JPEG RIGHT

GRADED TOP - UNGRADED BOTTOM

 

I uploaded these files for everyone to try them out if they want to:

103_0326 - 100ISO - https://mega.nz/#!DctRFaIY!eMeqsd-3incYu2IqBZQGW7zINRtwwF8aKO8AFd1gixQ

2592p15 - 3200ISO - https://mega.nz/#!PN0zQIIQ!7tcNWhfTOKDsUYvqZNocFb9bpnMyChdTamdawx5gbzI

3648p15 - 4000ISO - https://mega.nz/#!WdFHkAgZ!8MoMydWK55KXkOnDLR5AlwFslkC4PjZ2YDr2LKSN8KY

 

 

[pressland]

7 hours ago, SMGJohn said:

to say the least the difference is massive not just in the dynamic range but the pure colour

Wow, no kidding. So this could be achieved by using the MJPEG codec? I hope some clever people are on the case.

[SMGJohn]

42 minutes ago, pressland said:

Wow, no kidding. So this could be achieved by using the MJPEG codec? I hope some clever people are on the case.

Yes, Motion JPEG with high bitrate is far superior to HEVC and there is a Motion JPEG codec in the NX1 already for the 480p mode, Chant said he would look into it for me regarding the increase of resolution and possibly bitrate. 

It was a good move by Samsung to include the Motion JPEG codec besides the HEVC

[Marco Tecno]

Well, I guess this is not really like mjpg, in the sense that mjpg won't have such a high bitrate...anyway...a huge difference indeed.

[RieGo]

omg i never realised you get so much ugly sharpening, even though you did sharpness -10. seems like video postprocessing messes things really up.
so let's assume we get mjpeg codec - wouldn't it go through the same video processing? i think so. we definitely need a way to go around this horrible sharpening... 

OR it really is a sideeffect of hevc... i'm not sure.

well, ungraded looks way better (still too much) in terms of sharpness. did you do any post sharpening?

[SMGJohn]

5 hours ago, Marco Tecno said:

Well, I guess this is not really like mjpg, in the sense that mjpg won't have such a high bitrate...anyway...a huge difference indeed.

Depends to be honest, average JPEG in this sequence are 1,49Mb so that would translate to 22,35Mbps bitrate for 15fps which is hilariously tiny

If you pumped that up to 120fps for 1152p you get 178,8Mbps bitrate. Of course Samsung has given us JPEG compression so even at normal it still outperforms the HEVC, imao.

Now the 6480x4320 JPEG's at super fine compression has file sizes from anything between 4 to 15 from my own experience but lets say 15Mb then.

375Mbps bitrate for 25fps at the full output of the camera which is 6.5k at 3:2 aspect ratio, like the ol' television cameras, gonna need some BIG memory cards for this one or else you going to run out of space real quick but seeing how the bitrate is still quite below the limit of UHD-II speed and according to the Samsung mem dev test the Lexar 2000x can do 546.99Mbps average now whether this dev test is reliable or not is different question. But still it gives us a good estimate, now these are all rough poorly done maths but just to give us an idea what the camera could potentially do. We 'might' be looking at 50/60fps 6.5k MJPEG video here (750Mbps for 50fps and 900Mbps for 60fps with super fine compression), remember that MJPEG is barely processor heavy and at such extreme bitrates its laughable low CPU demanding, there might be other technicalities that would render this dream of 6.5k MJPEG video completely useless, but I suppose we just gotta wait to see what Chant might unveil in the holy grail for now.

3 hours ago, RieGo said:

omg i never realised you get so much ugly sharpening, even though you did sharpness -10. seems like video postprocessing messes things really up.
so let's assume we get mjpeg codec - wouldn't it go through the same video processing? i think so. we definitely need a way to go around this horrible sharpening... 

OR it really is a sideeffect of hevc... i'm not sure.

well, ungraded looks way better (still too much) in terms of sharpness. did you do any post sharpening?

Yes the JPEG have even less sharpening so there is indeed more in-camera sharpening happening behind closed doors. 

From what I can tell on the shitty 480p MJPEG mode it does not seem to suffer too much from sharpening but imao its 480p its low res so its hard to tell non the less. Hopefully someone turns off the in-camera sharpening, and yes I did do post sharpening but not a whole lot and both the HEVC and JPEG got the same amount of sharpening, this was done deliberately to show the effects of what sharpening does to HEVC compared to the JPEG and its drastically, the in-camera sharpening is bothersome to say the least. 

[BrorSvensson]

On 2016-03-26 at 3:33 PM, SMGJohn said:

Depends to be honest, average JPEG in this sequence are 1,49Mb so that would translate to 22,35Mbps bitrate for 15fps which is hilariously tiny

If you pumped that up to 120fps for 1152p you get 178,8Mbps bitrate. Of course Samsung has given us JPEG compression so even at normal it still outperforms the HEVC, imao.

Now the 6480x4320 JPEG's at super fine compression has file sizes from anything between 4 to 15 from my own experience but lets say 15Mb then.

375Mbps bitrate for 25fps at the full output of the camera which is 6.5k at 3:2 aspect ratio, like the ol' television cameras, gonna need some BIG memory cards for this one or else you going to run out of space real quick but seeing how the bitrate is still quite below the limit of UHD-II speed and according to the Samsung mem dev test the Lexar 2000x can do 546.99Mbps average now whether this dev test is reliable or not is different question. But still it gives us a good estimate, now these are all rough poorly done maths but just to give us an idea what the camera could potentially do. We 'might' be looking at 50/60fps 6.5k MJPEG video here (750Mbps for 50fps and 900Mbps for 60fps with super fine compression), remember that MJPEG is barely processor heavy and at such extreme bitrates its laughable low CPU demanding, there might be other technicalities that would render this dream of 6.5k MJPEG video completely useless, but I suppose we just gotta wait to see what Chant might unveil in the holy grail for now.

Yes the JPEG have even less sharpening so there is indeed more in-camera sharpening happening behind closed doors. 

From what I can tell on the shitty 480p MJPEG mode it does not seem to suffer too much from sharpening but imao its 480p its low res so its hard to tell non the less. Hopefully someone turns off the in-camera sharpening, and yes I did do post sharpening but not a whole lot and both the HEVC and JPEG got the same amount of sharpening, this was done deliberately to show the effects of what sharpening does to HEVC compared to the JPEG and its drastically, the in-camera sharpening is bothersome to say the least. 

22mbps is 180 mbit/s so thats pretty big.

[SMGJohn]

1 hour ago, BrorSvensson said:

22mbps is 180 mbit/s so thats pretty big.

You are mistaking bits and bytes.

One byte = 8 bits

1 megabit = 0.125 megabytes

180 megabytes = 1440 megabits

 

So in order to find out either you simply multiply bytes with 8 to find out the bits, in order to find out the opposite simply divide by 8, and remember to follow the metric prefixes for most accurate results https://en.wikipedia.org/wiki/Metric_prefix

17 000 000 bytes * 8 bits = 136 000 000 bits

17 000 000 bits / 8 bits = 2 125 000 bytes

17 000 000 bits * 0,125 bytes = 2 125 000 bytes

 

You can also use Google for quick conversions, just type into Google 4326 megabytes to megabits and you get the answer 4326 megabytes = 34 608 megabits etc

[BrorSvensson]

45 minutes ago, SMGJohn said:

You are mistaking bits and bytes.

One byte = 8 bits

1 megabit = 0.125 megabytes

180 megabytes = 1440 megabits

 

So in order to find out either you simply multiply bytes with 8 to find out the bits, in order to find out the opposite simply divide by 8, and remember to follow the metric prefixes for most accurate results https://en.wikipedia.org/wiki/Metric_prefix

17 000 000 bytes * 8 bits = 136 000 000 bits

17 000 000 bits / 8 bits = 2 125 000 bytes

17 000 000 bits * 0,125 bytes = 2 125 000 bytes

 

You can also use Google for quick conversions, just type into Google 4326 megabytes to megabits and you get the answer 4326 megabytes = 34 608 megabits etc

But in your post you say each photo is 1.49 mebabytes and for 15 fps you will get 22.35 megabytes/s which equals 180 megabit/s

[SMGJohn]

2 minutes ago, BrorSvensson said:

But in your post you say each photo is 1.49 mebabytes and for 15 fps you will get 22.35 megabytes/s which equals 180 megabit/s

Windows never displays bytes keep that in mind its always bits.

Mbps = Megabits per second

MB/s = Megabyts per second

[Syme]

39 minutes ago, SMGJohn said:

Windows never displays bytes keep that in mind its always bits.

Mbps = Megabits per second

MB/s = Megabyts per second

I've never seen a file manager report a file size in bits. It's always bytes. Computer hardware typically cannot even store or operate on a unit smaller than a byte independently. Megabits are almost exclusively used for data rates, not storage sizes.

Just to be sure I downloaded the files and checked them myself (on linux). Sure enough they are all at least 1 million bytes.

Therefore the correct calculation for the 2048x1152x15fps stream is 1.5*10^6 bytes/frame * 8 bits/byte * 15 frames/s = 180*10^6 bit/s, which is approximately 180Mbit/s.

Which is exactly what one would expect from typical JPEG compression ratios.

[SMGJohn]

EDIT: Guess I should not use other forums for what Windows displays and what not, my mistake

[Otto K]

Syme is right. 

Edit - to clear things up - file managers will report sizes in base 10 (kilo is 10^3=1000) or in base 2 (kilo is 2^10=1024) but always in bytes, never in bits. 

[SMGJohn]

41 minutes ago, Syme said:

I've never seen a file manager report a file size in bits. It's always bytes. Computer hardware typically cannot even store or operate on a unit smaller than a byte independently. Megabits are almost exclusively used for data rates, not storage sizes.

Just to be sure I downloaded the files and checked them myself (on linux). Sure enough they are all at least 1 million bytes.

Therefore the correct calculation for the 2048x1152x15fps stream is 1.5*10^6 bytes/frame * 8 bits/byte * 15 frames/s = 180*10^6 bit/s, which is approximately 180Mbit/s.

Which is exactly what one would expect from typical JPEG compression ratios.

 

16 minutes ago, Otto K said:

Syme is right. 

Edit - to clear things up - file managers will report sizes in base 10 (kilo is 10^3=1000) or in base 2 (kilo is 2^10=1024) but always in bytes, never in bits. 

My bad then, learn new things everyday, I should just have used a calculator to begin with and not messing around with my terrible maths again.

http://toolstud.io/video/filesize.php?imagewidth=2048&imageheight=1152&framerate=60&timeduration=1&timeunit=minutes