Why does my video play back with blotchy pixels when brightness is added?

[usalabs]

Some time a go I bought Corel Video Studio Pro X5 mainly because it's easy to use, so, here's the problem.

I've been appointed to be a local band's sound/video engineer, and recently recorded the band playing at a local venue, but the lighting wasn't that good, (rotating colored lights, dim ceiling lights, etc etc, but in the video, it's light enough to just make out the band members, so, I wanted to slightly lighten the video, but applying the 'Brightness' filter and selecting a reasonable setting, when playing the video, the video has severe blotchy greys and whites as if the entire video is at a very low resolution. These are the current details of the video:-


NTSC drop frame (29.97 fps)
MPEG files
24 bits, 720 x 480, 29.97 fps
Upper Field First
(DVD-NTSC), 16:9
Video data rate: Variable (Max. 8000 kbps)
LPCM Audio, 48000 Hz, Stereo

How do I stop the blotchy areas, but still have a decent brightness?

[jagabo]

upload a short sample

[budwzr]

Those are quantization artifacts.

[manono]

How do I stop the blotchy areas, but still have a decent brightness?

For one thing you don't wreck the black levels by turning blacks to grey. That's what you do if you make a simple black levels adjustment, which I suppose is what Corel's Brightness filter does. As jagabo says, though, a sample will tell the tale.

[usalabs]

Being that I have no reason (except to upload a sample file here), to split a video, I have absolutely no idea how to cut 100MB of video using Corel Video Studio, I've even tried using a free video cutter but it doesn't support the mpg file that Corel exports to, so the only way I can think of is to upload the unfiltered and filtered videos (1.45GB each) to youtube, and post the url in here.

[poisondeathray]

Is it a mpeg2 program stream ? (.mpg or .mpeg) ?

Use mpeg2cut2 or dgindex to cut a sample

[sanlyn]

Oops, poisondeathray beat me to it. Oh, well, for what it's worth....

Utube is useless for our purposes and will ruin your video. You say your video is mpg. It would be best to have something of the original, unprocessed video. But a few seconds of sample MPG can be cut without harm using the free DGIndex, which is part of the DGMPGdec utility. It's simple to use, and you need only the .m2v that it cuts out for you (an .m2v is an MPG without the audio). Get DGIndex here: https://www.videohelp.com/tools/DGMPGDec

If you need help with it, just ask.

[usalabs]

I downloaded DGIndex and loaded up the mpg file into it, and set the start and end markers, but there is nowhere in any of the menus to export the video between the markers, assuming that the video between the markers is less than 100MB (max file size for files in this forum).

Addendum.
This bit of info might be helpful.

The camcorder the video was recorded on, saved it to DVD format (VOB), and I had to use a free VOB->MPG convertor for Corel to import it.

Simply renaming .vob to .mpg wouldn't work, as Corel didn't recognize the format, unless it was converted first.

[poisondeathray]

Open the vob directly in dgindex , mark in/ mark out

File=>save project & demux video . This will give you .m2v elementary video and audio

Upload the .m2v

[usalabs]

ok, I found the .m2v file and uploading it,, it's 16MB

Here's the link for the sample:-

http://files.videohelp.com/u/216927/sample_video_RAW.demuxed.m2v

I'm also preparing another sample with the brightness filter applied.

[sanlyn]

Thanks for the sample. Wow. At a stingy 2934 kbps average bitrate, fast action, fast camera movement, zooms, low light.....it's anyone's guess how you'll clean up all the compression artifacts and block noise.

[manono]

The black levels are fine, so my own speculation about what's wrong was off. However, the whites are blown out badly. This is the final product? The one you said had severe blotchy greys and whites as if the entire video is at a very low resolution? But it's pretty bright where I am now and I'll have to wait until dark to be able to see it better so some others may have more informed advice.

The bitrate and the quality is kind of low for the whole sample, so maybe budwzr's comment (Edit: and sanlyn's above) is more on the money. How much video did you try and stick in the DVD? How long is the video?

[usalabs]

no, that sample is RAW VOB straight from the DVD, no editing, I'm preparing another sample that uses the brightness filter.


http://files.videohelp.com/u/216927/Sample_2_with_filter.demuxed.m2v

The size of the entire video is 1.09GB and 00:27:58.04 in length,,, the mini DVD is 4.7GB plenty of space,,, the video settings on the camera were set to HQ, with low mic reference to stop clipping.

EDIT:

I had a decision of using either slow shutter, infra-red, or normal, but as slow shutter/fast color would show a ghost trail when moving, and infra-red would show everyone green, the only other option was normal setting.

[manono]

Kind of too bad you captured the performance as DVD to begin with.

Oh, that's the original? After being brightened it's going to look pretty bad.

There are much better ways to make it easier to see. Also, if you stuck 3 hours of interlaced concert footage in a single DVD5, you're kind of doomed to begin with.

If you're going to be their video guy, you'll have to learn how to do this right.

[usalabs]

Kind of too bad you captured the performance as DVD to begin with.

The camcorder ($900 worth) only has 2 options, DVD or Memory stick, I suppose I could set it on a tripod, hooked up to a laptop USB or Firewire, on a table, but it has to be in a place where no-one walks around or in front of.

But, in a crowded bar that's kinda impossible.

[sanlyn]

Understood about the location. A shoulder brace would have helped. Waving the camera around like a baton doesn't work well, even in broad daylight. You have bright light on figures in the forefront (they'll be washed out...nothing you can do about that). And you could heighten the midtones somewhat, leaving darks and brights alone, because the crushed darks and blown out brights don't have any detail anyway. But it's the low bitrate that really kills it.

Corel Video Studio can't help much with this.

[usalabs]

I tried a tripod, but, people walking in front of it, and then you get the tipsies that sorta just float and bash into people, it's pretty hard to keep the camcorder steady, even with a tripod, I even tried in the DJ booth where the PA's and stuff are, which is raised above the floor, but the shielded glass has semi circular bars that get in the way, so the only position is in front of the staging area, but off to the right of that area is the smoking area, and people are walking in front of the stage to get to the smoking area, which makes it harder to film the band.

[sanlyn]

Understand that. The logistics are often a nightmare. Sometimes you might need a little of your own light somewhere. But better light and better loc won't much better with a low 3-hour bitrate. That's far too low for fast movement.

[NoBuddy]

Your problem is your camera. Recording with such a low bitrate, directly to DVD - and in such difficult locations: no way to get it nice.
Check out, if you can borrow another camera and try it again. Editng your mpeg file from DVD will be a nightmare.
And the band will blame you...

[usalabs]

I just looked at the camcorder, it doesn't show what bitrate it can record at, but it does show that it can record at 9Mbs HQ, which I believe is DVD5 max rate.

[sanlyn]

I managed to get rid of some of the noise with MCTemporalDenoise, but industrial strength filters like that on low bitrate video will cut some sharpness from it. Even if the blockies get cleaned up somewhat, very dark and very bright areas will show banding on re-encode. The interlace is kiinda noisy too, I think QTGMC and Vinverse can fix that and give you a smoother look. It'll take a lot of processing, though, in lossless AVI.

[manono]

I just looked at the camcorder, it doesn't show what bitrate it can record at, but it does show that it can record at 9Mbs HQ...

That would be much better. Especially if you're going to be editing and filtering it later on, you want to capture at the highest bitrate/quality possible.

Judging from that first sample, you used a much lower setting.

...which I believe is DVD5 max rate.

Not exactly. Read here for the DVD specs:

https://www.videohelp.com/dvd

[usalabs]

Nope, I was wrong, further digging into the cams settings found I can record at a max of 17Mbs with a resolution of 1440x1080 but only if switched to memory stick (using 16GB), DVD setting can max at 9Mbs

[poisondeathray]

That's too bad.

Is this a 1 night event? - can you record another performance ?

(Or do you have a time machine ?)



Many camcorders that record variable bitrates will lower the recording bitrate during low exposures. If you can set it, force CBR

An HD recording will most likely look better - and even if it had the same problems, you can denoise and it will still look better than the SD recording when scaled down to the same SD frame size because it's oversampled . I'm guessing the 1440x1080 17Mb/s is probably AVCHD

[jagabo]

Budwzr was right. You have quantization errors -- block artifacts from MPEG 2 compression. Basically, the MPEG 2 encoder decided you didn't need to see details in the dark areas and eliminated them. You can run a deblocking filter, adjust the gamma, levels and saturation, and add a temporal smoother. But it's never going to look good.

The deblocking filter will reduce the blocks but the smooth gradients left behind will look blurry. Bringing up the gamma will make dark detail more visible but the blockiness and noise will be more visible too. Temporal smoothing will smooth out some of the interframe noise but low contrast moving details will get obliterated.

[budwzr]

... so maybe budwzr's comment (Edit: and sanlyn's above) is more on the money.

I just threw that out as a general "catch-all" big buzz-phrase. I can always adjust the comment to fit any scenario.

[usalabs]

That's too bad.

Is this a 1 night event? - can you record another performance ?

(Or do you have a time machine ?)

Sorry no time machine

Yes, it was a 1 night event, they maybe playing the same venue next month, they don't know yet.

I can record a better performance at the band's practice room (a rebuilt garage turned into a band room, with audio recording facilities).

[poisondeathray]

I can record a better performance at the band's practice room (a rebuilt garage turned into a band room, with audio recording facilities).

And add some brighter lighting , record in HD at the highest bitrate mode




Another "odd" thing is this sample is 30p, field shifted - i.e. not interlace content. Very odd for a camera to record that

[sanlyn]

Another "odd" thing is this sample is 30p, field shifted - i.e. not interlace content. Very odd for a camera to record that

Ah-ha. So that's why I had such a tough time with combing effects. They wouldn't go away until I deinterlaced completely. Should not have had to do that.

I'm with budwzr and jagabo. Quantization problems + stingy bitrate = bad video.

The deblocking filter will reduce the blocks but the smooth gradients left behind will look blurry. Bringing up the gamma will make dark detail more visible but the blockiness and noise will be more visible too. Temporal smoothing will smooth out some of the interframe noise but low contrast moving details will get obliterated.

The gradients are both blurry and blocky, and many of them change shape occasionally. Note the distorted ceiling detail and banding and crud in shadows on right-hand wall. Not enough data bits to maintain motion, either. And true, if you raise the darks between RGB 20 and RGB 50 or so, there's no detail down there, just mostly grainy blur. You can't raise gamma without washing out the midtones that do exist in the brighter people and instruments..

The attached used a bitrate 2X the original. After most of the noise is gone, all that's left is an almost posterized lack of detail. It's not just a case of being a "dark" picture. If you remove even more of the noise seen below, it looks worse.

[budwzr]

Another "odd" thing is this sample is 30p, field shifted - i.e. not interlace content. Very odd for a camera to record that

Hmmm... me smell-um something Apple.