Name
Abstract | ||
|---|---|---|
Traditional clustering methods are inherently difficult to handle with a large scale of images, since it is expensive to store all the data and to make pairwise comparison of high-dimensional vectors. To solve this problem, we propose a novel Binary K-means for accurate image clustering. After hashing the data into binary codes, the weights assigned to the binary data are based on the global information and the weights for the binary centers are adapted iteratively. Then, in each iteration, with the center-adaptive weights the distance between the binary data and the binary centers is computed by the weighted Hamming distance. As the data and centers are presented in binary, we can build a hash table to speed up the comparison. We evaluate the proposed method on three large datasets and the experiments show that, the proposed method can achieve a good clustering performance with small storage and efficient computation. |
Year | DOI | Venue |
|---|---|---|
2017 | 10.1007/978-3-319-77383-4_40 | ADVANCES IN MULTIMEDIA INFORMATION PROCESSING - PCM 2017, PT II |
Keywords | Field | DocType |
Hashing,Image clustering,Center-adaptive weights,Weighted Hamming distance | k-means clustering,Pattern recognition,Computer science,Binary code,Hamming distance,Hash function,Artificial intelligence,Binary data,Cluster analysis,Binary number,Hash table | Conference |
Volume | ISSN | ISBN |
10736 | 0302-9743 | 9783319773827 |
Citations | PageRank | References |
0 | 0.34 | 10 |
Authors | ||
3 | ||
Name | Order | Citations | PageRank |
|---|---|---|---|
| Yinhe Lan | 1 | 0 | 0.34 |
| Zhenyu Weng | 2 | 6 | 3.85 |
| Zhu Yuesheng | 3 | 112 | 39.21 |