It is very often that I have to do video motion tracking for interactive video campaign in my daily work.
Click here to see the motion tracking demo (give it a bit of time to fully load).
What I'm doing in this demo is drawing a video in a canvas, and using the corner pin tracking data to draw an image on top of the video.
I made 2 different videos, you can switch between them using the "switch video" button.
In the first video, I get the current video frame number by calculation (current time by video fps). And in the second video, I get the frame number using a different technique that is know to be quite accurate when we do it in Flash, I call that a binary counter.
The idea behind the binary counter is to export the video with some white and black rectangle that describe graphically the real frame number of the video. We've been using this technique in Flash for a while in the advertising industry, and Zeh Fernando talked about it last year.
You can see the binary counter on the right border of the video. The code gets a pixel in each bit of the binary counter (16 bits), find out if the bit is white or black to get a 0 or 1 and calculate a number out of it.
While you might not see a difference between the two frame number calculations at first glance, seeking the video with the binary counter version (using the top-right blue control) will keep the image and the video "synchronised". Pretty cool.
This version is not perfect compared to what we do in Flash, some frames still seem to be a bit "out". I also couldn't find a way to solve "pixel snapping drawing" in a canvas, in other words, being able to draw an image on half a pixel.
I just found out that what I call a binary counter is actually a VITC, which I didn't know! Interesting!