Did you take a look at the still that has CC at the end of it? I applied the famous "S" curve and added some secondary color correction as well as some guassian blur to the red channel to keep it from pixelating.

Thanks for the suggestion MH.

j razz

I assume this is footage from your HC1. I had to correct an entire wedding reception that was even more underexposed than this, all from an HC1. Not good at all in low light.

You have issues beyond just the correction needed.

If you can post about 10-15 seconds of this in the form of an m2t file, so I can work with the original, I'll see what I can do. The easiest way to do this is simply to capture from tape for fifteen seconds, and then upload the m2t to a free ftp site like yousendit.com.

Generally I do the correction with a combination of custom color curves and the levels tool. I then do color correction following the old BB tutorial, more or less. Then, the real secret is to get rid of the noise, which with your camera is unfortunately going to be an issue. For that I use the FFT filter than I've written about before.

Actually it is from an A1u, but the same none the less.

Here is the file (59MB):

footage & stills

j razz

I can't download the m2t because the file name has a space rather than an underscore character. I'm trying to find my ftp program (I usually just ftp with Explorer, and it is choking on this). Any chance you can rename the "Booth Clip.m2t" file so it has an underscore (i.e., "Booth_Clip.m2t")?

Sorry about that John. I normally do that, but just didn't think about it on this one. It now has the underscore.

j razz

My whack at it...
[url=http://glennchan.info/Proofs/forums/Booth_Choir.jpg]

.veg:
[url=http://glennchan.info/Proofs/forums/Booth_Choir_3.veg]

One thing you can try is masking the people in the ?pews? and making them darker to re-compose the image... this puts more attention onto the stage.

Well, maybe it wasn't the underscore. I still get an error message when I try to download:


I can download the image files without any problem. I don't know whether this is something on your side or mine. Perhaps it has to do with the size of the file.

Sorry I won't be able to help.

Firefox 2.0 downloads the .m2t as a 1.6KB file...

Baaad upload.

John, Course,

I think what it is is that my host does not recognize the file. Here is what I did, I changed the extension to .mpg. After you download it, just switch it back to .M2T.

I tried it and it worked after I made the change. I got the same error you did before I switched the extension.

j razz

Hi Glenn,

I like the masking idea. Is there a way to keyframe the mask so that when I zoom or pan it will either fade to the bottom (zoom) or move smoothly to the left or right depending on which way I am panning?

Also, the second picture that shows cc, I really like how that came out. Thanks for taking a "whack at it".

j razz

I'll have a VEG for you in a moment, and then I'll upload an m2t that has the FFT noise reduction. You can use the VEG on your original to see how far that gets you, and then compare with the FFT noise-reduced version.

Be another few minutes ...

Thanks John. I will not be able to look at it until this afternoon as I am not at my office. Is this FFT similar to Mike Crash's noise reduction filter?

Thanks again John. I will take a look once I get in this afternoon.

j razz

Is this FFT similar to Mike Crash's noise reduction filter?

Yes and no. Yes in the sense that it is designed to get rid of random noise. No because it does so in a completely different way. Noise can be reduced by blending adjacent pixels (spatial de-noising); blending pixels in the same location but from adjacent frames (temporal de-noise, like the Crash filter); and blending pixels by converting each frame into "frequencies" using an FFT, eliminating the high frequencies, and then converting back to the time domain. This is much less easy to understand, but it produces vastly superior results to the other two. If you have used any temporal de-noiser, including the Crash filter, you will find that it introduces a sort of "screen door" effect as the camera pans or as objects move. On slow moving scenes, it does a wonderful job without causing these problems.

You'll see shortly and be able to decide yourself.

Thanks John. I look foward to seeing what you have done.

j razz

OK, here's the VEG file I used to brighten the clip (link good for seven days):

VEG file

I turned on the Videoscopes and enabled the external display to a calibrated monitor. I then used the Levels fX to increase the gamma, while keeping the "baseline" of the blacks at the same point. I also reduced the highlights to 100 to avoid problems with the blown-out light fixtures.

I then, gingerly used the Curves fX to tweak the shadows a bit. You have to very careful with this fX because you can easily solarize your video and make it look very unnatural. However, it can help bring just a tad more detail to the shadows.

I then used the Color Corrector to slightly diminish the green cast.

Thus, if you use this VEG on the original m2t you sent, you should end up with the same video that I created.

I then took that video and used the FFT filter (inside of AVISynth) along with a chroma de-noiser, to get rid of the noise that you see crawling along the walls in the background and on the shoulders of the people in the foreground. I suggest that you open your video using the VEG I sent above, and then put this video immediately below. Put your original, uncorrected m2t on a third track. Loop this while monitoring on an external monitor. The differences should be immediately obvious. What you will see if you A/B the corrected clip with the corrected AND de-noised clip is that the de-noised clip will at first appear just a little softer. However, in most cases, if you look to see if a detail is missing, what you may see instead is that the detail is still there. What happens is that noise often gives you the illusion of the video having more detail, because the noise dots impart a certain hardness and sharpness. I am not saying that there isn't a small loss of detail, but I think you will be very hard-pressed to see it. What's more, the dramatic reduction in noise sure makes it easier to watch. Pay particular attention to the pulpit. I think it illustrates the differences pretty well.

Here's the link to the 60 MByte gamma corrected and de-noised m2t video:

Denoised, Chroma noise, Levels, Curves, Color Corrected

Link good for seven days.

Thanks again John. After viewing it this evening I will get back with you on it. I might shoot an email your way concerning the fft filter and the settings you used. I assume that it comes with avisynth or can be found on the avisynth page: http://avisynth.org/

j razz

I might shoot an email your way concerning the fft filter and the settings you used. I assume that it comes with avisynth or can be found on the avisynth page: avisynth.org/

Yes, it works with AVISynth, although it is a separate download. If you get interested in pursuing that, you can search for FFT in these forums under my user name, or I can provide you with a short tutorial. Of course, you may not like the results, so see if you think it is worth it.

Also, I corrected the VEG link in my last post -- I had the HTML code screwed up.

Here's the FFT script I used:

#Serve YUY2 from Vegas and convert back to RGB32 in this script if going into Mainconcept MPEG encoder or back into Vegas.

#Check the RGB 16-235 box in the external Mainconcept encoder.



loadPlugin("c:\Program Files\AviSynth 2.5\plugins\CNR\Cnr2.dll")

loadPlugin("c:\Program Files\AviSynth 2.5\plugins\fft3dfilter.dll")



#Modify this to point to the video file you use.

source=AVISource("D:\Frameserver.avi")

source1=assumetff(source)

chroma=source1.Cnr2("oxx",8,16,191,100,255,32,255,false)

fixed=chroma.fft3dfilter(sigma=8, plane=0, sharpen=1, interlaced=true)

final=fixed.ConvertToRGB24(interlaced=true)

return final

I like the masking idea. Is there a way to keyframe the mask so that when I zoom or pan it will either fade to the bottom (zoom) or move smoothly to the left or right depending on which way I am panning?
You can:

A- Manually keyframe the mask. It doesn't take too long, but it's tedious. Remember that you can be pretty sloppy.
B- Motion track the mask (i.e. send in a feature request ;) ). Boris Red, After Effects are two programs I believe that has motion tracking. Tighter integration between Vegas and such programs would be an alternative (i.e. much like the audio integration with Sound Forge works well).

A quick & dirty alternative to the mask is to use the spotlight function in the bump map. Here's a VEG that includes all the previous corrections, but also adds a spotlight on the stage area. I spent less than ten seconds on this, so I'm sure you could tweak it to make it look better:

Spotlight on Stage

As you can see, Glenn's idea of changing the lighting intensity on the stage area is really a good one. It adds the stage lighting drama that is totally lacking in the original available light shot.

Here's the original video:



Now, here's the image after I applied levels, curves, color correction and noise reduction:



Finally, here is that image with the spotlight from the bump map applied. I intentionally let the fX blow out the shirt a little.

That's certainly a huge improvement.
Glad I looked at these images, I've been shooting a bit of footage under similar lighting conditions and I'm far from impressed with how the images turn out. Under not that low lighting these little HDV cameras produce remarkably soft images. I'm left wondering if I wouldn't have done better using a PD170.
It seems that it might have as much to do with the optics as anything, out wide and iris open seems to be the problem.

Bob.

Glad I looked at these images, I've been shooting a bit of footage under similar lighting conditions and I'm far from impressed with how the images turn out.

Bob, your profile says you are using the FX1 and Z1, and others as well. Are you disappointed with these two under low light? While I haven't yet filmed anything in conditions quite as low-light as what was done here, I have found that if I have a few minutes to set the exposure, the results are usually pretty good. I guess it all depends on your benchmark, and since I never used a VX2100, which I guess is the all-time champ in low light, I may not know what I'm missing.

Anyway, I'd be interested to know if your complaints are with either of the above cameras.