Re: video cameras and quality of image

[Tim Sullivan x5830 <sullivan@kenyon.edu>]

Brian Oliver writes:
> Could someone give me some advice on what kind of video cameras to  
> purchase for use with Video Point Software for the PC.  I have already 
> obtained an Intel Video Pro III capture board. > 


Gary Hemminger writes:
 
I would second this question because I'm finding that the chief 
problem with Video Point is the quality of the images.  The object 
is often too fuzzy to click on.  What video format works best? 

I write:

I'm not an expert on Video Point, but we do a lot of video at Kenyon (and I am
co-author of PhysVis, Kenyon's freeware program getting kinematic data from 
captured video clips, so have some expertise in digital video.) 

I can't think how the problem can be ascribed to Video Point itself, rather the
problem may be the current state of video. VHS tape is the worst (about
200x512 resolution). SVHS is better at about 400x512. Direct camera out is even
better (limited by the CCD pixel resolution and quality of optics).

So, TV can have a much lower resolution than computer graphics and the direct 
comparison in digital video systems makes this clear. 

Second, video capture hardware has a maximum resolution. The predecessor
to the Intel Video Pro was the Smart Video Recorder that had a max resolution
of 320X240. I don't know the Video Pro specs, but the best I have seen is new
cards (like the Miro DC20 or DC30) with 640x480 resolution. We use a card that
is no longer available, but has a 320x240 resolution.

At 320x240, we found that we had to make many changes to get good video data. Our
cameras are equipped with zoom lenses that have a pretty wide range of
adjustment and are mounted on supports that allow a lot of flexibility in
height above the floor and pointing direction. During an experiment, students 
are instructed to spend some time being sure to get the aim and zoom such as to
see all the relevant motion, but not have wasted space either. This keeps the
object is as large as possible and reduces the "graininess" of position
information by having the range covered by as many pixels as possible. If the
object in motion doesn't have a distinctive feature to click on, we add one
where feasible. We experiment with different colored balls to get the most
visible ones. We have found it necessary to use drapes behind experiments to
get a uniform background. The cursor in PhysVis can be switched from white to
black to give it the best visibility in a particular image. We also found that
additional lighting in the form of flood lamps helps in some situations. 

I'm a big proponent of digital video and we have gone whole hog with it here.
But it can involve some big changes to existing experiments.

Tim Sullivan
sullivan@kenyon.edu