100% Quality Factor MJpeg Settings (very low compression - class 10 required!)

Sorry I edited my post after realizing what settings you use while you where answering to my previous one. By the way I have 4:2:2 checked.
 
Great stuff!

I've got muddy chroma @ 84Mbit in this scene with these settings:
MJPEG E1 Q: 280
MJPEG E1 T: 14
MJPEG E2 Q: 226
MJPEG E2 T: 14
MJPEG E3 Q: 200
MJPEG E3 T: 14
MJPEG E4 Q: 170
MJPEG E4 T: 14
I'll test with your settings and rapport back!

81Mbit_MJPEG_frame.jpg
 
@SammySammy High Q values are not everything. If the Q value is too high, it will just jump down to the next. So for example, if your scene can't handle Q1=400, it'll jump to Q2=300, and so forth. But keep in mind it's likely that it's constantly jumping between Q1-Q4 values, and thus creating a certain bitrate. The tables are more important than the solely the bitrate in that it describes how the camera is seeing the footage.

From these little experiments, here is a theory in terms of stability...

(EDIT)

I think the reason people encounter stability issues with other settings is likely card speed coupled with not knowing their Q4/T4 threshold for their particular lenses. With the settings in this thread, you can set Q1 to 1000, and it still won't give a write error because Q4 is the stablizer. However, using such a high Q1 value will make the codec much more likely to ignore Q1 and just use Q2-Q4 ... because there is no scene where Q1=1000 can actually perform. This can be tested by simply setting all Q values = 1000 and trying to record. You'll get a write error for everything you point the camera at, even if it's dark.

Definitely pretty interesting stuff. I'm further refining the settings I've put up to see if there is anything left to squeeze out of it.
 
thanks for the detailed reply, so are you saying that the type of lens can have that big of impact on this whole thing?i understand it should, but how much..
 
Think about it this way.. it's detailed scenes that tax the codec. That being said, I just tried 422 with the thread settings... Q:400,300,250,188 / T:4,4,4,4 with no write errors on the 14-140 kit lens.

----
Precision=8 bits
Destination ID=0 (Luminance)
DQT, Row #0: 1 1 1 1 2 1 3 3
DQT, Row #1: 1 1 1 1 1 3 3 3
DQT, Row #2: 1 1 1 1 1 3 4 4
DQT, Row #3: 1 1 1 2 3 4 6 6
DQT, Row #4: 1 1 2 3 4 4 3 6
DQT, Row #5: 3 2 3 4 4 6 4 6
DQT, Row #6: 1 4 5 4 6 6 6 5
DQT, Row #7: 4 4 5 5 6 6 6 6
Approx quality factor = 97.07 (scaling=5.86 variance=3.18)
----
Precision=8 bits
Destination ID=1 (Chrominance)
DQT, Row #0: 1 1 1 1 6 6 6 6
DQT, Row #1: 1 1 3 3 6 6 6 6
DQT, Row #2: 1 3 4 6 6 6 6 6
DQT, Row #3: 1 6 6 6 6 6 6 6
DQT, Row #4: 4 6 6 6 6 6 6 6
DQT, Row #5: 6 6 6 6 6 6 6 6
DQT, Row #6: 6 6 6 6 6 6 6 6
DQT, Row #7: 6 6 6 6 6 6 6 6
Approx quality factor = 96.99 (scaling=6.02 variance=1.83)

Have to experiment some more to see if I can get it to generate an error. Might have to tick Q4 down to 187-6 for 422, but so far so good.

We might be able to lower Q1 without impacting the footage as well, which would mean a tighter spread between Q1-Q4... giving the camera possibility of utilizing a higher quality Q for complex footage.
 
Modulr I just tried your Q400, 300, 250, 188 / T4 setting with 4:2:0. It gives the best quality on "normal", "medium" scenes I saw so far but as expected movie recording of my high detailed scene canceled after a few seconds.

To make your setting work with my lens and 4:2:2 you say I only have to lower Q4 (128) until it works?
 
Yes. From my experience, Q4 is the last stop. If it can't maintain at Q4, you'll get a write error... so lower it until it works. Definitely interested in the results with your lens. :)
 
Thanks, using Q4 from my 1.95Q / T4 setting did the trick. My high detailed scene has now a quality factor of 95 (previously 85) and many more scenes have 100! The upper left corner of the quantisation tables is "1" all the time. My congratulations and thanks for the hard work you've put into this :dankk2:

Stable setting with 20mm, 1.7 Pancake lens
4:2:2 checked
Q: 400, 300, 250, 179
T: 4, 4, 4, 4


1278317122.jpg

A pretty flower in 100% quality, FILM MODE SMOOTH -1, -2, -1, -2
 
Last edited:
mimirsan said:
I just tried Martins setting and its pretty good...nice stable 70+ mbs with 4.2.2...I actually prefer the look from this setting.
Click to expand...

Yes I also thought it was pretty good but modulrs setting result in better quality tables and 4:2:2 is also possible. What do you mean with "look".
 
Martin Koch said:
Yes I also thought it was pretty good but modulrs setting result in better quality tables and 4:2:2 is also possible. What do you mean with "look".
Click to expand...

I mean by "look" that there is less visable noise with your setting. With Modulrs setting I get quite a bit of noise...indoors and out.
Im running yours with 4.2.2 on a GF1 right now.

I still think PappasArts setting is about as optimised as it can get (without 4.2.2).
I have tried these settings all on a GH1 & GF1 (Gf1s image is no where as detailed as the gh's but still a vast improvement than what it was)...its great that you guys are posting table settings...I have no clue what these tables mean! :2vrolijk_08:
Thanks!!!
 
mimirsan said:
With Modulrs setting I get quite a bit of noise...indoors and out.
Click to expand...

Interesting, because better JPEG compression shouldn't influence the source image or introduce noise. But wait you don't have control over ISO on a GF1 don't you? Do you see this difference also on the GH1 in manual mode with same lighting and same scene?

As for table values it's actually simple: you want them as low as possible. :2vrolijk_08:
 
Last edited:
modulr said:
@SammySammy High Q values are not everything. If the Q value is too high, it will just jump down to the next. So for example, if your scene can't handle Q1=400, it'll jump to Q2=300, and so forth. But keep in mind it's likely that it's constantly jumping between Q1-Q4 values, and thus creating a certain bitrate. The tables are more important than the solely the bitrate in that it describes how the camera is seeing the footage.

From these little experiments, here is a theory in terms of stability...

(EDIT)

I think the reason people encounter stability issues with other settings is likely card speed coupled with not knowing their Q4/T4 threshold for their particular lenses. With the settings in this thread, you can set Q1 to 1000, and it still won't give a write error because Q4 is the stablizer. However, using such a high Q1 value will make the codec much more likely to ignore Q1 and just use Q2-Q4 ... because there is no scene where Q1=1000 can actually perform. This can be tested by simply setting all Q values = 1000 and trying to record. You'll get a write error for everything you point the camera at, even if it's dark.

Definitely pretty interesting stuff. I'm further refining the settings I've put up to see if there is anything left to squeeze out of it.
Click to expand...

Thanks for that post. That helps clear some of the confusion I had about the Q settings.

I shot our local 4th of July fire works event with your settings and the 4:2:2 color option turned on. It worked perfectly. There were no write errors but also no in camera playback. The footage is stunning.
 
Martin Koch said:
Interesting, because better JPEG compression shouldn't influence the source image or introduce noise. But wait you don't have control over ISO on a GF1 don't you? Do you see this difference also on the GH1 in manual mode with same lighting and same scene?

As for table values it's actually simple: you want them as low as possible. :2vrolijk_08:
Click to expand...

Ive just retested on both gh1 & gf1 with your updated settings....fantastic results (i might of forgot to lock the iso on the gf originally)
 
mimirsan said:
Ive just retested on both gh1 & gf1 with your updated settings....fantastic results (i might of forgot to lock the iso on the gf originally)
Click to expand...

Great I'm also convinced and will use the modulr settings from now on.
Modulr settings - doesn't that sound good too?
 
Definitely good to hear the settings are working so well for everyone! And really good to know 422 is an reliable option. BTW, awesome shot Martin! not in a million years did I expect the camera to be able to output footage at this level of detail.

Even better news, I'm testing some new numbers that are yeilding even better results for complex scenes. The complex scene being some little bamboo bushes outside my house... lots of tiny leaves and patterns.

Original settings...

Q:400,300,250,188
T:4,4,4,4
----
Precision=8 bits
Destination ID=0 (Luminance)
DQT, Row #0: 1 1 1 1 3 2 6 6
DQT, Row #1: 1 1 1 1 2 5 6 6
DQT, Row #2: 1 1 2 2 2 5 8 7
DQT, Row #3: 1 2 2 3 6 7 10 10
DQT, Row #4: 1 2 3 6 7 8 6 11
DQT, Row #5: 5 3 6 7 8 10 8 10
DQT, Row #6: 2 7 9 8 10 11 10 9
DQT, Row #7: 7 8 9 9 11 11 10 10
Approx quality factor = 95.26 (scaling=9.48 variance=7.14)
----
Precision=8 bits
Destination ID=1 (Chrominance)
DQT, Row #0: 2 1 2 2 10 10 10 10
DQT, Row #1: 2 2 5 6 10 10 10 10
DQT, Row #2: 2 5 7 10 10 10 10 10
DQT, Row #3: 2 10 10 10 10 10 10 10
DQT, Row #4: 7 10 10 10 10 10 10 10
DQT, Row #5: 10 10 10 10 10 10 10 10
DQT, Row #6: 10 10 10 10 10 10 10 10
DQT, Row #7: 10 10 10 10 10 10 10 10
Approx quality factor = 94.92 (scaling=10.16 variance=4.80)

Compression stats:
Compression Ratio: 4.76:1
Bits per pixel: 5.05:1





Next (same shot)...

Q:500,325,250,188
T:4,4,4,4
----
Precision=8 bits
Destination ID=0 (Luminance)
DQT, Row #0: 1 1 1 1 1 1 2 2
DQT, Row #1: 1 1 1 1 1 1 2 2
DQT, Row #2: 1 1 1 1 1 1 2 2
DQT, Row #3: 1 1 1 1 2 2 3 3
DQT, Row #4: 1 1 1 2 2 2 2 3
DQT, Row #5: 1 1 2 2 2 3 2 3
DQT, Row #6: 1 2 3 2 3 3 3 3
DQT, Row #7: 2 2 3 3 3 3 3 3
Approx quality factor = 98.01 (scaling=3.98 variance=3.69)
----
Precision=8 bits
Destination ID=1 (Chrominance)
DQT, Row #0: 1 1 1 1 3 3 3 3
DQT, Row #1: 1 1 1 2 3 3 3 3
DQT, Row #2: 1 1 2 3 3 3 3 3
DQT, Row #3: 1 3 3 3 3 3 3 3
DQT, Row #4: 2 3 3 3 3 3 3 3
DQT, Row #5: 3 3 3 3 3 3 3 3
DQT, Row #6: 3 3 3 3 3 3 3 3
DQT, Row #7: 3 3 3 3 3 3 3 3
Approx quality factor = 98.39 (scaling=3.23 variance=0.48)

Compression stats:
Compression Ratio: 3.10:1
Bits per pixel: 7.74:1







and finally (same shot)!

Q:600,330,260,188
T:4,4,4,4
----
Precision=8 bits
Destination ID=0 (Luminance)
DQT, Row #0: 1 1 1 1 1 1 1 1
DQT, Row #1: 1 1 1 1 1 1 1 1
DQT, Row #2: 1 1 1 1 1 1 1 1
DQT, Row #3: 1 1 1 1 1 1 1 1
DQT, Row #4: 1 1 1 1 1 1 1 2
DQT, Row #5: 1 1 1 1 1 1 1 1
DQT, Row #6: 1 1 1 1 1 2 1 1
DQT, Row #7: 1 1 1 1 2 2 1 1
Approx quality factor = 98.47 (scaling=3.06 variance=5.95)
----
Precision=8 bits
Destination ID=1 (Chrominance)
DQT, Row #0: 1 1 1 1 1 1 1 1
DQT, Row #1: 1 1 1 1 1 1 1 1
DQT, Row #2: 1 1 1 1 1 1 1 1
DQT, Row #3: 1 1 1 1 1 1 1 1
DQT, Row #4: 1 1 1 1 1 1 1 1
DQT, Row #5: 1 1 1 1 1 1 1 1
DQT, Row #6: 1 1 1 1 1 1 1 1
DQT, Row #7: 1 1 1 1 1 1 1 1
Approx quality factor = 100.00 (scaling=1.54 variance=1.58)

Compression stats:
Compression Ratio: 2.56:1
Bits per pixel: 9.37:1

That is the lowest compression and highest bits/pixel I've seen yet. I haven't switched on the 422 for this setting, because I want to shoot some more to see if it's consistent, reliable, and not doing anything else weird.. but wow, this camera is INSANE! still a little unsure that such a wide spread between Q1-Q4 is a good thing, we'll see soon enough... heh.

So if anyone is feeling a bit rogue and adventurous, give it a try... :)

Q: 600,330,260,188
T: 4,4,4,4
 
Last edited:
Just tried it with 4:2:2 and the tables are better than ever.


NEW MODULR SETTINGS with lower Q4
Q: 600, 330, 260, 179
T: 4, 4, 4, 4
Low detailed scene: 100 / 73.96 mbps / PLAYBACK NO / MOVIE OK
High detailed scene: 97.69 / 90.94 mbps / PLAYBACK NO / MOVIE OK
4:2:2
Clip recording time display: 1m45s


But I also noticed that it doesn't keep these tables throughout the clip. I recorded my high detailed scene for about two seconds.

Tables of first second or so
Code: 
Precision=8 bits
  Destination ID=0 (Luminance)
    DQT, Row #0:   1   1   1   1   1   1   2   2 
    DQT, Row #1:   1   1   1   1   1   2   2   2 
    DQT, Row #2:   1   1   1   1   1   2   3   3 
    DQT, Row #3:   1   1   1   1   2   3   4   4 
    DQT, Row #4:   1   1   1   2   3   3   2   4 
    DQT, Row #5:   2   1   2   3   3   4   3   4 
    DQT, Row #6:   1   3   4   3   4   4   4   4 
    DQT, Row #7:   3   3   4   4   4   4   4   4 
    Approx quality factor = 97.69 (scaling=4.61 variance=3.08)
  ----
  Precision=8 bits
  Destination ID=1 (Chrominance)
    DQT, Row #0:   1   1   1   1   4   4   4   4 
    DQT, Row #1:   1   1   2   2   4   4   4   4 
    DQT, Row #2:   1   2   3   4   4   4   4   4 
    DQT, Row #3:   1   4   4   4   4   4   4   4 
    DQT, Row #4:   3   4   4   4   4   4   4   4 
    DQT, Row #5:   4   4   4   4   4   4   4   4 
    DQT, Row #6:   4   4   4   4   4   4   4   4 
    DQT, Row #7:   4   4   4   4   4   4   4   4 
    Approx quality factor = 97.90 (scaling=4.21 variance=0.76)

Table of a frame near the end.
Code: 
Precision=8 bits
  Destination ID=0 (Luminance)
    DQT, Row #0:   7   7   7   7  22  15  36  36 
    DQT, Row #1:   7   7   7   7  15  29  36  36 
    DQT, Row #2:   7   7  15  15  15  29  51  44 
    DQT, Row #3:   7  15  15  22  36  44  65  65 
    DQT, Row #4:   7  15  22  36  44  51  36  73 
    DQT, Row #5:  29  22  36  44  51  65  51  65 
    DQT, Row #6:  15  44  58  51  65  73  65  58 
    DQT, Row #7:  44  51  58  58  73  73  65  65 
    Approx quality factor = 68.68 (scaling=62.65 variance=266.78)
  ----
  Precision=8 bits
  Destination ID=1 (Chrominance)
    DQT, Row #0:  15   7  15  15  65  65  65  65 
    DQT, Row #1:  15  15  29  36  65  65  65  65 
    DQT, Row #2:  15  29  44  65  65  65  65  65 
    DQT, Row #3:  15  65  65  65  65  65  65  65 
    DQT, Row #4:  44  65  65  65  65  65  65  65 
    DQT, Row #5:  65  65  65  65  65  65  65  65 
    DQT, Row #6:  65  65  65  65  65  65  65  65 
    DQT, Row #7:  65  65  65  65  65  65  65  65 
    Approx quality factor = 66.78 (scaling=66.45 variance=154.02)

Same static scene no camera movement no change in lighting.

Unfortunately JPEGSnoop doesn't work well with the movie files but I helped myself by cutting out a few frames near the end with QuickTime Pro and saving this as a self contained movie without any recompression.
We haven't looked any further than the first frame so far and I suppose the camera jumps to Q4 very soon with such high bitrates.
Another real life scene showed a "degradation" over time from 97 to 92 quality with the old settings from this morning.

We have to test this "degradation" with real life scenes. I would love to see that it's minor.
 
Last edited:
modulr,
thanks for your hard work!
I just tried your setting and it worked fine but I was able to playback the clip using VLC player, not qucktime player. I guess it's normal?





modulr said:
Thanks all!

BTW, these settings are basic 720p and only require setting the MJPEG quality and table values. I have not explored 420->422 nor running 1440x1080. I'll experiment with those to see how far things can get pushed.
Click to expand...
 
modulr said:
That is the lowest compression and highest bits/pixel I've seen yet. I haven't switched on the 422 for this setting, because I want to shoot some more to see if it's consistent, reliable, and not doing anything else weird.. but wow, this camera is INSANE! still a little unsure that such a wide spread between Q1-Q4 is a good thing, we'll see soon enough... heh.

So if anyone is feeling a bit rogue and adventurous, give it a try... :)

Q: 600,330,260,188
T: 4,4,4,4
Click to expand...


I am very impressed with this and have tried your last stable setting, so now I will have to try this one..

One question I have, are the screen grabs you are showing in the begining of the thread just the .jpg files that are captured as you start the recording? If so, what are you setting the camera to or what lens are you using, because with my kit lens or my Nikkors I am not getting things that clean... yours look like they came out of my Nikno D3 they are so crisp... would love to hear more about that.

Thanks,
Cory
 
Martin Koch said:
Just tried it with 4:2:2 and the tables are better than ever.


NEW MODULR SETTINGS with lower Q4
Q: 600, 330, 260, 179
T: 4, 4, 4, 4
Low detailed scene: 100 / 73.96 mbps / PLAYBACK NO / MOVIE OK
High detailed scene: 97.69 / 90.94 mbps / PLAYBACK NO / MOVIE OK
4:2:2
Clip recording time display: 1m45s
Click to expand...

Martin,

I have a new codec test chart put together- which is way better at crash testing the codec then the star chart I put up- The key is to shoot it bright/contrasty off the monitor- None of my extreme settings have survived this chart. This is a great to fine tune your numbers as well see how far the codec will go in Mbits.

Here is a link to the 2K chart: http://www.sendspace.com/file/5f0if7

I also posted a sample picture to see the chart. - However download the 2K version from the above link.

MIKOS • MLPappas
http://TWITTER.COM/PAPPASARTS
http://PAPPASARTS.WORDPRESS.COM
http://MIKOSarts.wordpress.com
http://MIKOSarts.wordpress.com/page-2-of-blog
http://MIKOSarts.COM
 

Attachments

  • new codec tests chart PAPPASARTS.jpg
    new codec tests chart PAPPASARTS.jpg
    341.8 KB · Views: 0
Last edited:
I just tried mjpeg with 400 300 250 188 / 4 4 4 4 using Papas chart and Sandisk 16 extreme failed after a couple of seconds.

PappasArts said:
Martin,

I have a new codec test chart put together- which is way better at crash testing the codec then the star chart I put up- The key is to shoot it bright/contrasty off the monitor- None of my extreme settings have survived this chart. This is a great to fine tune your numbers as well see how far the codec will go in Mbits.

Here is a link to the 2K chart: http://www.sendspace.com/file/5f0if7

I also posted a sample picture to see the chart. - However download the 2K version from the above link.

Pappas
Click to expand...
 
You must log in or register to reply here.
Back
Top