Compressing huge AVI files from low quality concert VHS captures

[jeff442]

Nice to meet you all. I have been reluctantly thrown into the world of video capture and conversion. I have collection of ~45 concert bootlegs on VHS. I would like to ultimately archive these and share (probably via some type of torrent). Based on recommendations from this site, I purchaed the Canopus ADVC-100. My first attempt at capturing (audience shot 1hr. clip) resulted in a 12 gig avi file! The quality is awesome (no worse than the source, which is all you can ask). My question is where to go from here. Any software I can use to clean the video up? More importantly, how do I get the file to a shareable size. I downloaded the "Super" conversion program, but no matter what setting I try, the resulting .mpg is blocky or significantly worse quality than the original. I find it hard to believe that I high quality 480p motion picture can fit on a 4.7gb DVD, but a 1hr VHS of questionable quality can't be stored in a manageable size. Am I using the wrong settings? How small can I make this file without a noticeable decrease in quality? Thanks in advance for your help... I am very exicted to be part of this community.

[CogoSWSDS]

Download Virtual Dub from virtualdub.org and then get FlaXen VHS (it's a Virtual Dub filter) from Neuron2.net to clean up VHS. Warning: FlaXen VHS is slow, you can get faster filtering by using Convolution3D but that requires AVISynth (available from AVISynth.org) and doesn't do quite as good a job as FlaXen VHS, IMHO. These will clean up video somewhat and make compression with fewer macroblocks easier.

You could also try 2-pass VBR compression - takes twice as long though. 1st pass is spent doing analysis on the video to determine which scenes need more bitrate and which need less bit rate. 2nd pass is spent doing the actual encoding.

CogoSWSDS

[Richard_G]

What you have to remember is that your Canopus is generating a DV.AVI file, exactly the same as you get from a digital camcorder. Although this is compressed it is complete in that every frame is stored complete. DVD video uses MPEG compression which only stores a complete frame (I frame) every 15 frames, the ones in between are information stating what has changed since the last I frame. DV.AVI is just over 13Gb per hour, converting to DVD compliant MPEG reduces this down to around 4.5Gb per hour with minimal noticable loss of quality.

No matter what you do, compressing the file will result in a loss of quality. The compression throws away some of the information. What you have to do is decide what type of compression you intend to use, what frame resolution and how much compression you an get away with before the quality suffers too much.

[solarfox]

One reason they can get a two hours-plus feature-length movie onto a DVD with high quality is because they're starting with a high-quality source.

The encoder can't really tell the difference between "noise" and "motion" -- all it knows is that it sees pixels changing from one frame to the next, so noisy source video causes the encoder to waste a lot of bit rate on trying to encode the noise.

My process is generally something like this:

(1) Capture the video into an AVI file. (This will be in DV form, since I'm also using an analog-to-DV converter.)

(2) Use VirtualDub to de-noise the video. The exact combination of filters I use tends to vary depending on the source material and how much noise reduction I need; some recordings are worse than others, and some kinds of video (such as cartoons) can stand heavier levels of noise reduction that would make live-action footage look unnaturally flat and "plastic"-like. Sometimes, I'll make a one- or two-minute "test clip" from the original source and then try several different filter combinations on it, to see which one looks best. Since I don't have a DV codec that VDub can use to encode with, the de-noised video is written out to a new AVI file using the lossless HuffYUV codec.

(3) Use TMPGenc to encode the denoised video to DVD-ready MPEG2. For VHS captures, I'll typically set the frame size to half-D1 (352 x 480), since the horizontal resolution of VHS isn't all that great anyway -- and by halving the number of pixels to be encoded, it makes it easier for the encoder to get the image into the desired bit rate without macroblocking. I also set it to use 2-pass encoding, and to save the audio and video as separate Elemental Streams rather than as a single multiplexed MPEG since DVDLab (my DVD authoring program of choice) prefers demultiplexed ES's.

Now, some may wonder why I don't just frameserve straight into TMPGenc instead of wasting disk space with an "unnecessary" intermediary file. Well, there's a good reason for that: Because those denoising filters are s...l...o...w, and if you're doing 2-pass encoding, that means that the denoising process has to be done twice -- once for the analysis pass, then again for the encoding pass. This can add up to a whole lot of wasted time if you've got a lot of stuff to be processed...

[JaredKnight]

Richard_G and solarfox - two very informative posts...I was not aware of the 1 every 15 frame (MPEG)..thank you