BGP Face Retrieval Based on Coding Pyramid

摘要

The traditional face image retrieval method is to compare the target picture with all the pictures in the database one by one, resulting in great time consumption. In this paper, the cascaded binary gradient pattern (BGP) is used to hierar-chically encode faces and form an encoding pyramid. The method of using fuzzy matching to exact matching is used when searching, according to the coding level, and the idea is based on the coarse-to-fine search. It is verified in the Yale Library that the method significantly shortens retrieval time and improves search efficiency.