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

[NoBuddy]

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

Where did you get this from?
The file I downloaded from post #10 (supposed to be the original) says:

Frame rate : 29.970 fps
Standard : NTSC
Color space : YUV
Chroma subsampling : 4:2:0
Bit depth : 8 bits
Scan type : Interlaced
Scan order : Top Field First

[_Al_]

It is weird that Corel Video Studio got it so bad anyway, compare it with Vegas encode, same bitrate, and the same encode but denoised with Neat video, just to compare, no levels adjusted (to make it brighter) not sure if it is good idea
http://files.videohelp.com/u/198160/no_denoise.mpg
http://files.videohelp.com/u/198160/neat_video.mpg

[manono]

Where did you get this from?
The file I downloaded from post #10 (supposed to be the original) says:

Frame rate : 29.970 fps
Standard : NTSC
Color space : YUV
Chroma subsampling : 4:2:0
Bit depth : 8 bits
Scan type : Interlaced
Scan order : Top Field First

There's no contradiction there. MediaInfo says how it was encoded, not what the content is like. In addition, field-shifting makes it appear as if it really is interlaced. However, if you bob the video, each bobbed field has a duplicate, the proof that the source is really progressive and not interlaced.

Look at it this way. Where the capital letter is the top field and the small letter is the bottom field, a progressive sequence might go like this:

ABCDEFGHIJ
abcdefghij

Each top field is from the same point in time as its bottom field and together they make up one progressive frame. With field-shifting you get this:

ABCDEFGHI
bcdefghij

This makes it appear to be interlaced, but realigning the fields (a simple matter with an AviSynth field-matching filter) makes it progressive again.

Good find, pdr.

[sanlyn]

Are we talking "realigning" fields, or "matching" fields? Field matching I know about. But what about realignment? I didn't think they were the same thing.

Serious question. Looking at it, the combing doesn't look like genuine interlace. It looks like a sideways misalignment.

[poisondeathray]

Are we talking "realigning" fields, or "matching" fields? Field matching I know about. But what about realignment? I didn't think they were the same thing.

Serious question. Looking at it, the combing doesn't look like genuine interlace. It looks like a misalignment of identical images. I can imagine that QTGMC(InputType=1) or QTGMC().SelectEven() would give similar results, but that's deinterlacing, not realigning.

It's essentially the same thing in this case, either field matching or manually realigning

In avisynth, either

field matching

TFM()

or

manually realigning

SeparateFields()
Trim(1,0)
Weave()

The benefit of using TFM() is if the pattern changes (for whatever reason), it should automatically adapt , at the expense of slightly slower speed. In both cases , the fields should match up now giving you full progressive frames, but the 2nd method will be offset compared to the 1st method . The 1st method will start with a missing field, so it will be deinterlaced (you can check with TFM(display=true) to give you a full frame

There is residual combing eitherway, so you probably have to add something like vinverse. Or it might give you better results using some of QTGMC's processing, but in general I try to avoid deinterlacing progressive footage

[usalabs]

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.

Sorry to disappoint, but it's not Apple cam or software, the cam is a Sony HDR-HC9 and the software is Corel Video Studio Pro X5, running on an HP Pavilion p6-2143w, 8GB RAM, quad core APU, (not to be confused with CPU), Windows 7 Ultimate, and 1TB HD.

I tried to convince the band to pull together to get a Sony HDR-AX2000, but at $3,500 it may take a very long time to get one

[sanlyn]

Thanks again, pdr. I'll give both a try and keep this clip around as another learning experience. QTGMC + Vinverse still had bad combing, even with InputType=1. SelectEven() was the cure, but I'd rather take your tip and avoid deinterlacing this type of source when possible.

[jagabo]

There is residual combing eitherway

Yes. It's not real combing (ie, a frame made from two non-matching fields) but a result of different quantizers used for the matching fields. As mentioned VinVerse() can get rid of that.

Also, because of the severe over compression, there are some vertical comb-like artifacts. Those can be smoothed away with TurnRight().VinVerse().TurnLeft(). An alternative to the paired VinVerse() sequence is Blur(1.0).Sharpen(0.7); if you can live with the slightly blurred result.

[sanlyn]

This time, at least, combing and some other junk cleared using the suggested TFM and SeparateFields+Trim method. Surprising, but no telling how another clip might react. The attached mpg used SeparateFields, allowing me to cut the same denoising filters' power by half. If nothing else, the attached demo shows that raising the darkest regions gets very little. Just nothing down there, except noise and banding. I used the same filters as the previous post, with NeatVideo turned almost off (used mostly to reduce simmering low-level schmutz that dances around in the dark during camera pans). Could remove even more noise, and even some banding -- but there wouldn't be any video left to watch. I removed the previous effort posted yesterday.

Thanks for the tips on that realignment gltch. You learn something new every day in this joint.

[usalabs]

Thanks guys for the awesome help.

I burnt a DVD-RW to show the band and they accepted the fact that the venue lighting was crap, and the venue itself was so-so, they said to scrap it, (luckily it was on a DVD-RW) but they are playing at a different venue next week, which gave me the chance to check it out.

The lighting is a lot better, and there is an area specifically designed for photography/video where people are not allowed during a performance, it has a platform that can be raised or lowered a few feet, to get the best shot(s).

Now, that is something I can work with, and this time, I 'll set the cam to use a memory stick at the highest rate possible (17Mbs), and as they are only playing a set of 5 (or 6 if they have time for 1 more) songs, which takes about 30 to 40 minutes of video.

When I've added an intro title and ending credits, and maybe polish the video a bit with some scene cutting and transition effects, I'll post a sample here for you guys to comment, and make any suggestions.

@sanlyn I was looking at your second attempt sample, I didn't notice this, but look at the soloist during his solo, it looks like his right arm has a skin disease, lol

[sanlyn]

Good luck.

[jagabo]

Low light performance requires big glass (think telescopes). Big glass generally means big sensor and big camera.

[sanlyn]

And along with those requirements, you'll need more $$$.

[usalabs]

Well, Here's a new update.

I finally recorded the band with excellent vid quality, great lighting and excellent audio quality, but,,,, (and yes there is always a but), no matter how I try I can't convert the vob from the cam dvd to a lossless avi format for editing, without banding (interleaving), WMP plays the dvd great, no banding whatsoever, but when I import the vob into Adobe Premiere Pro, CS5.5 and preview it, there's a lot of banding, the same when importing into Corel VideoStudio Pro, so I tried using Cinelerra under Ubuntu Linux, but this time the audio is way out of sync and still a lot of banding.

I'm assuming that vob files are actually mpeg format, so theoretically, I should be able to just rename the extension from .vob to .mpg and edit it as an mpeg, but being that mpeg is a video compression method, I would rather convert it to a lossless format, I would prefer the cam to record in raw format dv, but the cam converts the video into dvd vob format before writing it to the disk.

[jagabo]

I guess the banding you are talking about is interlace comb artifacts. If you deinterlace you will lose half the temporal resolution and at least some of the spacial resolution (depending on how well the deinterlacer works).

You can't just rename a VOB to convert it to an MPG container. VOB2MPG or Mpg2Cut2 can remux VOB to MPG.

Converting from VOB or MPG to losslessly compressed AVI will not improve the quality. In fact, it will lose superblacks and superbrights if you don't do it correctly. It might make it a little easier to edit though. Be sure your editor knows the imported video is interlaced. DV is not lossless.

[usalabs]

I tried VOB2MPG it did the job good for remuxing, and I had a smooth video in MPG format.

After adding a beginning intro title, some transitions, as well as the ending credits, a preview showed great,,,, no interlace comb artifacts

Another problem now exists, as the video was recorded in 16:9, and imported into either Corel VideoStudio or Adobe Premiere Pro, the previews show everyone like matchsticks because the default import took it to 4:3, with black bars outside the left and right of the video, even if I changed the project default aspect ratio to 16:9, the preview shows everyone squashed, top and bottom, but it is full screen.

Why can't video editing software use the source aspect ratio when importing, EG, import a 4:3 video, and work on it as 4:3, or import a 16:9 and edit it as 16:9? instead of the software developers assuming the that all videos are 4:3.

Later when I record the band again, this time I'm using my laptop fire-wire connection to the cam and record raw data direct from the cam, non-compressed, if audio can be recorded raw, then later after editing can be saved as mp3, then video should be able to be recorded raw and later after editing saved compressed to fit on a DVD.

[jagabo]

In most editors you can right click on a video in the timeline and set various source properties there.