Conclusion

A system was developed that is able to track the ball, draw a red cross at the ball's estimated position on the video monitor, and hit the ball. This project is turned out to be tightly constrained by the timing of the video frame grabber (vertical retrace interrupt). Therefore it ended up with essentially a single periodic task.
Some possible improvements are:

• A variable resolution algorithm might be used to improve the timing. It could change the zoom factor from 8 to 1 by zooming to the region where the ball is expected. In other words, it would start with a low resolution, but finally read the pixels with high resolution only in the small area where the ball is supposed to be. It would start to read pixels with 8 to 1 zoom, and if it finds the ball, then it would change the zoom to be 4 and read pixels from a quarter of the frame, find the ball again, and change the zoom to be 2 and so on.
• A digital filtering algorithm might solve some of the noise problem.
• Use of a faster bus and processor would allow this system to have more time to execute the tracking algorithm. The faster bus would speed up reading the pixels, which is currently taking most of the period, and the faster processor would allow doing more other processing.

Finally, if there were time for more complicated processing, this might allow putting several balls on the table at once, predicting bounces, and planning shots to optimize trajectory.