Preparing your source footage

[Chris Northcross]

Hey again guys! I'd like to hear some input from folks who use SA to auto-rotoscope video. Specifically, do you prep your source footage before you start working in SA at all (run a color correct or levels operation)? Most of the footage I've used in the past has been with a Panasonic HVX 200 and the colors tended to be rather rich so I didn't bother much with prep work. I shot some footage recently on a cheap handy cam and am finding I that I have to finesse it a bit before I run it through Studio Artist to define my edges a bit better. I'd like to hear workflow tips and ideas if you have them.

[Synthetik]

Like you said, it depends on the characteristics of your original footage and how you plan to process it. having good contrast and edge definition in your original footage always is a good thing, so if that's not the case then doing some preprocessing to boost that can be a good thing. I use the image compressor and the blur ip op with an edge composite for contrast boost and edge enhancement respectively.

[Victor Ingrassia]

Generally enhancing via ADJUST, combined with an ImOp like SMOOTH or even Rank Arae Filter. But it really depends on the nature of the source footage as well as what you a re planning to do with it.

~v

[David Kaplan]

this may seem obvious, but if you are using interlaced video it really helps to de-interlace it before using it as a source for studio artist work. you can do this very well via compressor.

[Synthetik]

You can use the Deinterlace Ip Op to do this as well.

[David Kaplan]

yes you can, that's the quickest way; you're basically dropping one interlace field and losing half your image resolution - which for most SA work is probably just fine. going the other route will take much longer but provide sharper end results depending on your settings, if this matters to your workflow. and there are many other programs to do this as well.

[Synthetik]

You could build a PASeq in 3.5 that incorporates a write frame command as an action step halfway through the sequence, so you could process one interlace field, dump the result for processing that field as a frame write, and then process the other interlace field using a repeat of your processing steps. All in a single PASeq.

The fact that you can record write frame steps in a PASeq is a little known feature of 3.5. It's actually an example of something that has not been incorporated into v4 at this point.