Feature Extraction For Biological Images

One of the most difficult problems in biological imaging is extraction and classification of similar features taken from different perspectives. Through the use of the cross ratio, this problem has been successfully treated for the case of four points on a line. Unfortunately, features that cannot be described by straight lines do not have this property of unchanging curva ture, and require more complex techniques. These techniques will be very powerful tools in feature based image registration, in particular for use in dental radiographs. In this paper we will discuss an elegant method to perform this recognition and classification based upon the principles of projective invariants.