Recently I've been using a program called Dxtory to capture footage while I play battlefield 3 such that I can review my playing and possibly arrange compilations of excellent kills etc. The problem is that these avi files are huge (~320 GB/hr) and long(1 hour)
Basically the solution I envisioned was to automatically detect kills and get a list of times at which they occur so that I could quickly find all the good footage without having to watch the whole video. I've achieved much of that goal but I'm hoping to improve the process.
Currently the steps are as follows in premiere.
The end result is that I get very close to 1 frame from every 4 seconds of the original recording at 60 fps with the relevant image data aligned to the upper left corner. Example images shown here. Original and After Processing
After I've generated the series of png files, I analyze them with a MATLAB script I wrote that returns a list containing the time and frame number when the kill text appears.
Is just a straight forward method for just grabbing every 240th frame rather than the convoluted method I use above. I ask mainly because this method drifts away from the proper frame over time. It's not too bad but I feel like this is a pretty straight forward task that I could easily do in MATLAB but for the enormous filesize.
Secondly, is there a way to crop down to just the relevant area without all the extra black space so I'm exporting images that are more like 400x100 rather than 1920 x 1080? I cant seem to figure out how to do this.
Finally, after getting a list of times and frame numbers, is there any way to use that to automatically extract clips with a script such that I can delete the wasted space? Perhaps there is another program more suited to this kind of task that you could recommend for me?
Thanks for your time.