- Indexing System for Pictorial Database: Using Framework of Image Recognition
- No.36, p.23-44
A model of the framework of the human image recognition was built, and the indexing of pictorial database which applied it was devised. A series of experiments in which subjects assigned terms to the images was carried out, and a model consisted of image recognition categories such as “cente”, “space and locatio”, “central and detailed descriptio”, etc. was built. On the basis of this model, a pictorial database was developed using the indexing which assigned index terms according to the categories. An experiment was carried out which compared the performance of this method with that of the PDL method. The overall performance of this method as measured by recall and precision was higher than that of the PDL method, showing the validity of the model and the applicability of the indexing method.