关键词: 信息处理技术, 局部嵌入算法, 可变k近邻, 数据降维

Abstract: In content based image retrieval, computation complexity will increase seriously as the demension of the image feature increases. When reducing the vector dimension with locally linear embedding algorithm, the number k of nearest neighbors should be determined. In this paper we propose a variable k neighbors LLE (VKLLE) method based on the distribution of the image feature. The VKLLE method can reduce the vectors dimension with keeping their original topology structure into a lower dimension space. Experiment shows that the proposed VKLLE method can achieve higher precision rate in content based image retrieval problem.

Key words: information processing, locally linear embedding, variable k nearest neighbor, data dimension reduction