A computational model for process-grammar

Shape analysis is a challenging topic in the research areas of human and machine vision. Recently, Leyton [1] proposed a process-grammar for analyzing morphological change of two-dimensional patterns. Because the process-grammar was designed to give a qualitative description of what has occurred in the intervening time, it does not quantitatively derive the later shape from the earlier one and the continuous family of "intermediate shapes" is therefore unspecified. A computational model is thus proposed to supplement the process-grammar for describing the continuous process-history of shape change. Based on the idea of elastic interpolation, the basic approach is to find a set of "forces" acting on one shape and trying to distort it to be like the other shape. The intermediate shapes generated by this process can be considered as the process-history of the two shapes. Five out of six rules in the process-grammar can be explained by the model without any modification. By a minor modification, the only rule remained can also be covered by the model.