C100: the internal noise reduction works with Ninja2

P

Pave1

Well-known member
I repeatedly heard here that the C100 camera doesn't output a signal via HDMI output with the applied internal NR.

In my latest test with the Ninja2 recorder this internal NR works (NR0, NR2, NR4):
http://www.videoproduce.cz/images/006b_NINJA2_NR0_at_21min_30sec_origQT.tif
http://www.videoproduce.cz/images/007b_NINJA2_NR2_at_22min_07sec_origQT.tif
http://www.videoproduce.cz/images/008b_NINJA2_NR4_at_22min_40sec_origQT.tif

https://vimeo.com/88598926

What do you think about this?

edit: it was with 4000 ISO, in video at 21:30, 22:07, 22:45
 
Last edited:
Thanks Pave1

Tangible differences between the frames - but not too much detail lost. Very useful.

Don't suppose you tried a similar test using WDR, did you?

Cheers,

Ben.
 
Yes, look at video at 23:27.

This test was made with 850ISO, 2000ISO and 4000ISO, with cine1 (cine1 with black gamma has brighter black level), wdr and log gamma at 25p and several times also at 25i. The 25i mode is really important for WDR gamma because there is much more noise - only in this case, no other gammas give these results, the external record is (only in this case) worse than internal. But all other gammas including wdr at 25p are ok, better with the external recoder (true 422 colors) and better also with the internal NR applied, the result is less noisier.

If you will download the orig H264 file (3GB) from vimeo then it is clear visible. From QT master it is a little sharper but there is not a big difference in this static type of the scene. In black parts are noise bigger, but color table at PIP effect also show the differences. It is the best to viewing this at a really dark room or at evening.

With the external recorder (true 422 colors with cca 185-200Mbps) you will get the best start point for advanced color grading, the black is free from the compression artifacts, there is only noise (with NR2 it is nice signal with a lot of details with acceptable level of noise).
 
Thanks, Pave1. I typically have internal NR set to zero in my picture settings. Do you have recommendations for acceptable levels of NR based on ISO?
 
drummondb said:
Thanks, Pave1. I typically have internal NR set to zero in my picture settings. Do you have recommendations for acceptable levels of NR based on ISO?
Click to expand...

If you will see the video (it is a dark place with poor light), then for example with 4000 ISO (tif pictures above) and for cine1 gamma is the NR2 nice, NR4 can be too much. So NR3 looks as the safe level of noise reduction. For 2000 ISO looks NR2 ok. Test this with wdr gamma (NR3-4 for 4000ISO) or for canon-log (NR 0-2 for 4000ISO can be nice).

I will recommend don't use unnecessarily high level of the NR (except for WDR in 25p where 4000ISO + NR4 is nice) for a very clean images because a little noise help with low 8bit color levels (256 shades) with prevention of banding (also the 185-200Mbps recording helps with this vs. uncompressed recording which doesn't help). If you present your video at web (with H264 codec with 8-12Mbps) then also this second re-compression clean some noise (so you can use a lower level of NR). This way of thinking helps to prevent using too much noise reduction.

If you use unnecessarily high level of the NR then on your editing monitor it looks ok, but after reexport there could be more artifacts especially in black parts. But level NR2 is safe for all purposes, with higher ISO (3600-4000) you can use NR3 (NR4 on your own risk .-) ) for externally recording.
 
....We don't want pictures without noise, we want pictures with an acceptable level of noise :-) .

(and if you test the noise level with canon-log gamma then you need to do it after color grading and I recommend to use the different level of ISOs than with others gamma curves)
 
Last edited:
Back when i was hearing it, i believe what people were saying wasn't that it didn't output USER controllable NR, but that it was noiser because aside from the NR settings that you can set, that there are more behind the scenes NR going on that we cannot adjust, which are not being applied when outputting, which makes sense, they assume people want the external output to have as little processing as possible.

nonetheless, good to see your results. The ninja star is darn tempting
 
Darren Levine said:
Back when i was hearing it, i believe what people were saying wasn't that it didn't output USER controllable NR, but that it was noiser because aside from the NR settings that you can set, that there are more behind the scenes NR going on that we cannot adjust, which are not being applied when outputting, which makes sense, they assume people want the external output to have as little processing as possible.

nonetheless, good to see your results. The ninja star is darn tempting
Click to expand...

Thanks. The Ninja Star will be at the end of May at Europe (I hope), I also want this magic lightweight box .-) !

I think that the less noise in AVCHD is the result of the bigger compression in H264, it is a general known issue for this sort of H264 (for example I saw differences between XF100 with mpeg2 long GOP and GH30 with AVCHD). Output to HDMI looks to be from the DIGIC DV III before the internal compression output part, the HDMI is uncompressed but signal procesors at video cameras usualy work with 422 color levels (there can be also 444 colors but I don't know anything about this in the C100's DSP), I think that this uncompressed signal is only made in output HDMI HW from 422 otput, so the 422 recorder make a lot of sense for me to get maximum from the C100 camera.

The C100 has the main "magic" denoiser at analogue domein (directly on the chip at early stage of processing), so I don't believe too much that there is some another "magic" about this, only differences at the compression type (and the CP noise reduction at Digic DV III which works for both output - HDMI and internal AVCHD). But I can be wrong .-) .
 
You must log in or register to reply here.
Back
Top