Chain Code Histograms for Rotation Invariance

A challenge for retrieval, recognition and representation of binary objects is to find invariants under different transformations. Rotation invariance allows for example to recognize a shape even if it is rotated. We proposed a method to obtain chain code histograms of skeleton and contour shape of objects. We used two representations: the Freeman chain code (F8) and directional Freeman chain code (AF8) to encode regular and irregular shapes. We compared these techniques and we obtained a sounds method to consider chain code as invariant under intermediate rotations.