Bilinear Discriminant Analysis Hashing: A Supervised Hashing Approach for High-Dimensional Data

详细信息    查看全文

• 刊名：Lecture Notes in Computer Science
• 出版年：2017
• 出版时间：2017
• 年：2017
• 卷：10115
• 期：1
• 页码：297-310
• 丛书名：Computer Vision ? ACCV 2016
• ISBN：978-3-319-54193-8
• 卷排序：10115

文摘

High-dimensional descriptors have been widely used in object recognition and image classification. How to quickly index high-dimensional data into binary codes is a challenging task which has attracted the attention of many researchers. Most existing hashing solutions for high-dimensional dataests are based on unsupervised schemes. On the other hand, existing supervised hashing methods cannot work well on high-dimensional datasets, as they consume too much time and memory to index high-dimensional data. In this paper, we propose a supervised hashing method Bilinear Discriminant Analysis Hashing (BDAH) to solve this problem. BDAH leverages supervised information according to the idea of Linear Discriminant Analysis (LDA), but adopts bilinear projection method. Bilinear projection needs two small matrices rather than one big matrix to project data so that the coding time and memory consumption are drastically reduced. We validate the proposed method on three datasets, and compare it to several state-of-the-art hashing schemes. The results show that our method can achieve comparable accuracy to the state-of-the-art supervised hashing schemes, while, however, cost much less time and memory. What’s more, our method outperforms unsupervised hashing methods in accuracy while achieving comparable time and memory consumption.