Project History
I once used a program that would take multiple images from a computer flat bed scanner and average them to eliminate some of the CCD generated noise in the image. A flat bed scanner works very good for this technique. As long as you make all the scans without lifting the cover each pixel should be over the same spot on the paper. The differnce in color value for this spot is the random noise. The idea is that the noise will not fall on the same pixel for each scan and by averaging the images you get a higher quality final image.

Trying the same technique for a VideoCamera
Obviously for this to work both the camera and the subject must be stationary, as well as the lighting conditions must not change. I wrote a program that would read a list of ppm formatted images average them and then spit out an image. Click on each image to see the full 640x480 image or view all of them Full Sized.

Single frame
Maximum Light
Same Light as Above
Less Light

8 frames averaged
30 frames averaged
47 frames averaged

Averaging images can clean up Video Camera images
I was surprised at how clear the images came out after being used to the image quality that comes out of that camera. It sure beats the X10 camera I have, but after seeing the these images I have a better appreciation for why home video equiptment doesn't compare to the quality shown on tv. Naturally if you want good quality still images go buy yourself a descent still digital camera and you'll do better than what I've shown here, but it was a fun project anyway.

Hardware used
The Video camera is a Sony Camera that is a few years old, it uses (analog) 8mm tapes. The tv capture card is a Pinnacle PCTV Pro tuner card that uses the Bt878 chipset. xawtv was used to capture the images.

Source is available.
The averating software was designed and written on Linux.

Download Directory

Written by David Fries <>

Valid HTML 4.0! Valid HTML 4.0!
My pgp key is available and the fingerprint follows. See for details.
pub 1024D/CB1EE8F0 2001-08-21 David D. Fries <>
Key fingerprint = 7079 F7EA D7EA 8E93 5B84 1900 008F 39D9 CB1E E8F0
sub 1024g/D9B8B029 2001-08-21