Precise and Efficient Retrieval of Captioned Images: The MARIE Project.

摘要: The MARIE project has explored knowledge-based information retrieval of captioned images the kind found in picture libraries and on Internet. It exploits idea that are easier to understand with context, especially descriptive text near them, but it also does image analysis. approach five parts: (1) find captions; (2) parse interpret (3) segment into regions homogeneous characteristics classify them; (4) correlate caption interpretation using focus; (5) optimize query execution at run time. emphasizes domain-independent methods for portability expense some performance, although domain specification is still required. Experiments show prototypes more accurate than simpler methods, task very challenging work needed. Its processing illustrated detail part an Internet World Wide Web page.