Noise Solution Testing

Alex Leith said:
Personally, though, it would make me very nervous shooting this way!
Click to expand...
Me, too. I want to see exactly what I´m doing. At last this is one of the big advantages that video has. One question about ND filters: If you use them, less light gets to the sensor an therefore more noise should appear, right?

Barry_Green said:
That's what the oft-requested -3dB Gain setting would do. But no, the HVX doesn't offer that.
Click to expand...

But is there any workaround by combining other settings for example?
 
Last edited:
I conducted a noise test. Where can I post the stills?

Details of the test:

I did a simple test lighting a scene in different ways (with one light, two lights and three lights). I shot each locked down shot in 1080 and 720p. The results are interesting.

Thanks!
 
I also have about two minutes worth of video I shot on a sunny day in a graveyard and at a factory. It's compressed for the web, but looks pretty good (it's 82 MB).

I uploaded it to www.video.google.com about a week ago. It's still not up. I should try to post it somewhere else if you guys want to see it.

I could also provide full resolution short clips (around 45MB each).

Let me know.
 
I went to image shack. My still files are twice as big as what's allowed. Are you sure you want me to compress the files? That may not be such a good idea.
 
Danilo, I received your PM and I will send you some links, as well, I will phone you as soon as possible in order to answer your doubts. Or as attemp to.
 
Emanuel,
Thanks for the reminder. I'm uploading 1 of the 10 stills now. I'm trying file factory. It's taking forever (over 5 minutes now) for just one still. I'm not sure how long I can take this!
 
thefilmaddict said:
Emanuel,
Thanks for the reminder. I'm uploading 1 of the 10 stills now. I'm trying file factory. It's taking forever (over 5 minutes now) for just one still. I'm not sure how long I can take this!
Click to expand...

Why not summarize for us what you found out while you're uploading. :-)
 
NO STILLS YET, BUT HERE'S WHAT I FOUND OUT:

NOISE TEST:
I did a camera noise test yesterday and made 10 stills showing the results. Here are the names of each file:

01a 720.tif
This is one omni light spotted on a white wall. I wanted to see how the camera would handle the areas of the shot that were not as bright as the center of the image. This was shot in 720p.

01b 1080.tif
Same as above, but in 1080.

02a 720.tif
This was a band of light. The bottom half is properly exposed. The top half is darker (though not completely black). This was shot in 720p.

02b 1080.tif
Same as above, but in 1080.

03a 720.tif
This is an evenly lit wall photo. I just wanted to see how 720 compared to 1080 overall.

03b 1080.tif
Same as above, but in 1080.

04a 720.tif
I shot 3 items. I exposed for the key light on those items. There is also a back light turned on. The wall to the left is not lit, but it's white, so lots of light is spilled on it. The room to the right is completely dark. This was shot in 720p.

04b 1080.tif
Same as above, but in 1080.

05a 720.tif
Same as above, but I added a background light. The lighting is more even all the way around.

05b 1080.tif
Same as above, but 1080.


WHY DID I DO THIS TEST:
I did this test because I wanted to see if shooting in 1080p reduced noise vs. shooting in 720p.

All of the images were exposed for the brightest part of the shot.

To be fair to the camera, the noise is not as pronounced when the image is playing/moving, but it's still there.


THE RESULTS (what I see/saw):
There doesn't seem to be a big difference between 1080 and 720p (if any). Perhaps the 1080 image is a little richer in chroma, but I do not see a big difference in noise (please tell me what you find). Notice in shots 5a and 5b where I lit the background. This more evenly lit shot is almost noise free.


FOR THE FUTURE:
I would like to be able to lower the noise and still shoot backgrounds that fall off or are completely dark. I hope we figure out how to do this. Perhaps we will find camera settings that will further reduce the noise.

If not, it would be nice to figure out how many stops lower we can let the background fall before the noise gets too loud. If anyone figures this out, please let me know. I will try to test that myself soon.


NOTE:
Someone came up with the idea to overexpose the image, then bring it down in post. I think this person suggested that you could overexpose by 3 stops. This did not work for me. I could not get detail out of things that were 3 stops over. Slightly overexposing the shot did not seem to reduce noise. Please prove me wrong if I did not test this theory properly.

I USED THE FOLLOWING SETTINGS IN THE TEST SHOTS LISTED ABOVE:
Detail Level -3
Detail Coring +7
Master Ped -2
Gamma Cinelike D
Matrix Norm
V Detail Freq Thin

The uncompressed TIFF Files were made in Final Cut Pro. The clips came from 2 projects (one project was 720p and the other was 1080). The images seemed to be scaled in 4x3. I did not know how to fix that, but it shouldn't change the results of this test.

DISCLAIMER:
I am not a rocket scientist. Don't take this info as gospel, it's just a simple test.
 
thefilmaddict said:
No luck posting the images. I will try again on Monday. These free sites are a pain in the @#&*)($ !@&!!!
Click to expand...
Try putfile.com

EDIT -- I've been used to posting stills without noticeable problems.
 
Last edited:
thefilmaddict said:
Someone came up with the idea to overexpose the image, then bring it down in post. I think this person suggested that you could overexpose by 3 stops. This did not work for me. I could not get detail out of things that were 3 stops over. Slightly overexposing the shot did not seem to reduce noise.
Click to expand...
What he did was OVERLIGHT by four stops (16x light level) and closed his iris four stops to achieve the equivalent "exposure." He then brought down the exposure by only three stops and pulled the highlights back out in post (+1 stop "overexposure").
 
Thanks NBC Shooter. That makes more sense. So what f-stop was he at? Did it help? Wouldn't doing this still introduce the same areas of gradient light (dark areas vs bright areas)? I can bounce a light off the ceiling to bring up the overall light level, then add a ton more light and light the way I want to, but the noise is still there. I guess the question is, how far did he let his background fall off vs how bright his subject was?

I need sleep. Did that make any sense?
 
Does anyone know if its true that the CCD cannot resolve fully above 5.6 - if this is true then overlighting and closing down isnt a reasonable solution
 
smelni said:
Does anyone know if its true that the CCD cannot resolve fully above 5.6 - if this is true then overlighting and closing down isnt a reasonable solution
Click to expand...

ND would solve the problem. Select your stop and use ND to bring down the light hitting the sensor to get your exposure.
 
I just found out that I have the day off. I would like to try the over lighting method. Just so I do it right, here's what I think the steps are (please tell me if I'm wrong):

1. To over light, I need the exposure to be three stops higher than the lowest f-stop on the HVX200. This means that my subject should be at 3 stops higher than 2.0, which would be 5.6. This is the exposure I will put my subject at.

2. This lighting method does not say anything about how far you can let your background lights fall off, so I will put some light on the background, but I will also leave some areas of the background unlit. I will try to light like I would to get a very moody look.

3. I will over expose the image by 1 stop. I will open the iris one stop to 4.

4. In post I will bring the video levels down to try to get the detail back in the over exposed places in the shot. I may also need to play with black and mid levels to get the image to look pleasing.

DO I HAVE THIS LIGHTING METHOD CORRECT? In theory, this method will decrease noise. I'll let you guys know what I find out.

Also, my current camera settings are:

Detail Level -2
Detail Coring +6
Chroma Level 0
Master Ped -1
Gamma Cinelike D
Matrix Norm

Should I change any settings for this test???
 
You must log in or register to reply here.
Back
Top