Video Editing Forums: Digital Director - View Single Post

Crusty · #2 (**permalink**) 01-18-2006, 10:26 PM

Preparing a video for net distribution isn't a straight-forward proposition, there are several things involved that some editing packages take care of auto-magically and others don't. I don't know what Premier does.

What I do depends on the source material and the desired output format. If the source is interlaced, like with MiniDV cams and analog Hi8, I deinterlace. The most important step. There are a several methods to do this. With 720x480 format, I usually resize to 320x240, this gets rid of the interlace and makes the output file much smaller. Anytime the vertical dimension is reduced by half the interlace is automatically eliminated.

If I want to retain the 720x480 format I choose the blend method. This is where field 1 and 2 are simply averaged together. This does cause blur on the vertical edges of moving objects, but there's always going to be some resolution loss and this seems to least noticable.

The next step is to smooth the video, since there are always some compression artifacts in the source. Mini-DV has the least, MPEG-2 has quite a bit. A slight blur filter works well or temporal smoothing - this blends a portion of previous frames into the current frame, this works because the distortion changes from frame to frame, but averages out in the eye. Since you'll be recompressing later, it's best to average this out beforehand, else the compression will magnify it and add it's own distortion, making a bad situation worse. The filtering also greatly improves the recompression efficiency, since you have only so many bits per second, there's no need to waste part of it on the distortion.

The next parameter is the compressor settings themselves. This is an art, one I've not mastered. Each compressor has it's own unique constellation of settings to fiddle with. One generic setting is constant bit rate or variable bit rate. Constant is fast and allocates the same number of bits per second to all frames, irregardless of how much complexity there is. VBR tries to allocate more bits per second to the busy scenes and less to the stationary scenes. This works best but you need to select how many passes it should make, one or two. Two is better than one, but does nearly double the time it takes.

Other parameters are motion sensing and quantization matrix stuff. Most compressors have defaults that are best, on average. I never diddle with these unless some expert has posted something that works better.

Now, choice of compressor, there's too many. I prefer Microsoft's WM9 codec. The distortion it produces tends to be spread evenly throughout the frame and not in nasty blocks like Mpeg 2 or Mpeg 4. All things considered, it's about the best. But, many video editors have no direct way to use them, so I use a stand-alone handler, Virtual-Dub. With it I can do all the deinterlacing, resizing, filtering and compression in one step. It can also handle the audio (re)compression. Plus it has a lot of other filters and adjustments. Frame rate decimation is a good way to conserve file size without giving up bery much quality.

The bad thing about virtual-dub is it only produces avi type container videos. This is a generic container and compatible with just about all PC media players, but there is no way to prevent copying. No digital rights management. If you need that, you'll have to use Microsoft's Media Encoder to make .wmv files and probably pay a licensing fee. Same for QuickTime .mov, the other popular container.