Last edited on 2005-01-17 13:57:51 by stolfi
Our multiscale fragment matching procedure was implemented as a function library and a set of programs, one for each major task -- outline extraction, filtering at various scales, conversion to curvature graphs, initial matching, candidate refinement, scale-to-scale matching, plotting, etc.. These programs were written in Modula-3 --- a very nice platform-independent systems programming language with Pascal-like syntax, which unfortunately never got the popularity it deserved.
Given the virtual demise of Modula-3, are currently in the process of converting all our programs to the C programming language -- specifically, for the Gnu/Linux platform, with gcc version 3.2 or higher. At present, only the information-content analysis package has been converted.
The source files are offered here "as is", without warranty, for whatever they are worth.
See Helena's dissertation [H99] as well as our papers and technical reports for further information on these programs.
Please note that these programs are research prototypes and not production tools. In partcular, there is no convenient documentation for the system as a whole. An overview of the programs and their use, in Portuguese and slightly outated, is included in Helena's thesis. Further information is provided by comments in the source files (especially in the Modula-3 library libm3pz) and by some test shell scripts that are included in the archive files. If you need to use or understand these programs, feel free to ask the authors for assistance.