Fast Multi-level Retrieval for Calligraphic Characters

To make full use of the digitized calligraphic works efficiently,a fast multi-level calligraphic retrieval approach is proposed,in which two levels of comparisons are used.For the first level,corner points and end points are extracted,and the features on them are compared,which can significantly reduce the number of comparisons of the pixels.For the second level,the result of the first level is used as input and the skeleton similarity based method is used.Thus the number of calligraphic characters to be compared is reduced significantly.Compared with the single level skeleton similarity based retrieval method,the proposed approach needs only half of the time,while the ratio between precision and recall is almost the same.It makes the retrieval for large scale calligraphic characters more efficient.