Thread Subject: harris features correlation

Dear all,
I've a view of a parking lot, I periodically apply harris detector to the image
and, where cars are present it find many points of interest.

I would like to answer the question : are the points in the same position of the image before or strongly similar?

I tryed the function matchbycorrelation.m available here
http://www.csse.uwa.edu.au/~pk/research/matlabfns/

 the autor says :
  Function generates putative matches between previously detected
  feature points in two images by looking for points that are maximally
  correlated with each other within windows surrounding each point.
  Only points that correlate most strongly with each other in *both*
  directions are returned.
  This is a simple-minded N^2 comparison.

but if I give to the function two similar images the function finds matches even in the different regions.. why?

ps. I've no problems in rotations or scale since the point of view is fixed.


thanks for answer, sorry for the question but I'm a beginner
Andrea