Video motion tracking and speed estimation

There is actually a very active listserver:

snipped-for-privacy@redhat.com

On a sunny day (Sun, 15 Mar 2009 17:54:06 -0500) it happened "Anthony Fremont" wrote in :

As you are using a composite camera, the signal is already split into Y and U +V for the color.

The idea of YUV is that you need less resolution for color vision, so many times the signal is transmitted as YUYV, or YUVYUV, where U and V each have 1/4 the resolution as Y. Anyways, there are many formats for that sort of encoding, but I have sort of standardised on the 'mjpegtools YUV format'.

That has a number of advantages. You know about Unix pipes, for example you can do: cat file.txt | grep 'Hello" where 'grep' is used as a filter to only get 'hello' through (if it is in the file.txt).

The mjpegtools project created a simple format to send YUV (color video of any format) from one application to the next.

This means you can do filtering, resizing, displaying, encoding, even sending via the net

YUV format (uncompressed) files are huge. But now we can do for example this, to send a H264 stream via the internet: camera | scaler | encoder | netcat

Or use a lowpass to get rid of noise: camera | filter | scaler | encoder > H264.avi

As mjpegtools yuv allows you to chain things, you can for example lowpass filter, de-interlace, resize, and even feed it back into a v4l device and have flashplayer use it so it is broadcast to many other users via some sites, for example for video conferencing.

There is a whole lot of tools in that package, xine, mplayer, and ffmpeg can read the format.

An other advantage of working with YUV in general is that you can process color and BW content separately.

last but not least I wrote 'subtitler-yuv':