A Rapid Quantization-based Image Retrieval Algorithm

Tingjia Shan

Department of Automation, University of Science and Technology of China, Hefei 230027, China

Qiang Ling *

Department of Automation, University of Science and Technology of China, Hefei 230027, China

Kaikai Song

Department of Automation, University of Science and Technology of China, Hefei 230027, China

Binbin Du

Department of Automation, University of Science and Technology of China, Hefei 230027, China

Feng Li

Department of Automation, University of Science and Technology of China, Hefei 230027, China

Song Wang

Department of Automation, University of Science and Technology of China, Hefei 230027, China

*Author to whom correspondence should be addressed.


Abstract

Fast image retrieval has been a fundamental problem in the area of image processing for a long time. This paper proposes a rapid image retrieval algorithm by improving the conventional nearest neighbor search through the implementation of vector product quantization and inverted indexing structure. Vector product quantization can efficiently accomplish the fast nearest neighbor search task, and has many great advantages in terms of storage requirements, retrieval speed and accuracy. In order to further reduce the search time, an approximate threshold-based distance estimation technique is introduced into the retrieval algorithm. Moreover, the quick sort method is implemented to reorder the image search results, which can significantly improve the performance of our retrieval algorithm.

Keywords: Product quantization, nearest neighbor search, image retrieval, inverted indexing structure


How to Cite

Shan, Tingjia, Qiang Ling, Kaikai Song, Binbin Du, Feng Li, and Song Wang. 2016. “A Rapid Quantization-Based Image Retrieval Algorithm”. Journal of Advances in Mathematics and Computer Science 18 (1):1-11. https://doi.org/10.9734/BJMCS/2016/27865.

Downloads

Download data is not yet available.