Match Moving Tutorial Part I

Match Moving. Big fancy word for extracting usable data from your video footage to create a camera in your 3d app for compositing.
Before we start I'll let you know I'm using Boujou. It's not free, but it's the most user-friendly tracking software I've found to date. You can use Icarus (free) or Autodesk MatchMover (free for students).
Button names may differ slightly but the concept is the same. So where do we start?

Footage
What kind of footage are you working with?
The majority of cameras out today are HD, atleast 1280x720. This gives you quite a bit of data to work with. Now we can talk all day about how many lines the camera actually resolves and how most HD I don't consider HD, but that's not what this tutorial is about. The key is always getting clean footage to work with.

-Here is another technical bit. With the movement from CCD sensors to CMOS sensors we've gotten a great improvement in video quality. But with new technology comes new problems, inevitably. Enter Rolling-Shutter. Again, I won't go deep into the technical stuff but rolling shutter is basically when one part of your frame is scanned horizontally instead of all at once. At worst this can make your footage completely un-trackable. People with Dslr cameras and FlipCams and cameras labeled CMOS sensor be warned.
Ways to avoid this are:
-Avoid terribly fast pans
-Stabilize your footage.
Now I don't mean you need a tripod because that defeats the purpose of what we are trying to do here. Keep your camera steady to an extent. Eg. Steadicam, gyros, jibs, or any means of getting fluid motion.
This will minimize or nearly eliminate any smearing or skewed images.

Here is an example of rolling shutter at just about it's absolute worst.

(Photo from wikipedia, not mine)

Here is a very important tip for capturing your footage. Include points of reference in your footage so your application can better re-create this space in 3d. This means, if possible, include tracking markers that you can later mask out in After effects. Another thing I learned is to include straight vertical and horizontal lines if at all possible. This will better help you're application know what tracking points are parallel or lie on X,Y,or Z planes.

So you have your footage, now what?
Well, the next step is to convert your video stream to an image sequence. You can do this with After Effects, Quicktime Pro, or I'm sure there is plenty of free software out there. Even if your tracking application accepts video formats such as MOV, AVI or WMV, you can run into codec reading issues (h.264, ProRes, Sorenson) that can be averted by converting to a JPEG or PNG sequence.


Part 2 in next post.