Character Recognition Using a Log-Autocorrelation Function

Chromosome analys is is complicated in tha t chromosomes of the same class appear in d i f f e r e n t shapes. We consider shape desc r ip t i on in terms of conic sec t ions . An i n d i v i d u a l chromosome is def ined as a non-negative func t ion f on the r e a l p lane, subject to c e r t a i n cons t ra in ts on p o s i t i o n , s i ze , o r i e n t a t i o n , e t c . A conic sect ion is represented by a quadrat ic form Q(x,y) 0, and a best conic fo r f is one fo r which 2 f ( x , y ) | Q ( x , y ) I dx dy is minimum. The problem may be formula ted as an eigenvalue problem of order s i x , the vector co r re sponding to the smallest root de f i n ing the best con ic . The chromosome shape is f i r s t described by f i v e basic parameters; w i t h these d i f f e r e n t chromosome shapes may be d is t ingu ished . Next a conic skeleton is def ined which r e f l e c t s the general shape of the denser par ts of the chromosome. The chromosome may then be analysed w i th the a i d of the ske le ton , in p a r t i c u l a r the skeleton suggests t ransformat ions which may p r o f i t a b l y be ca r r i ed out on the p a t t e r n .