By the virtue of BOF to describe high-dimensional data, in this article, we propose an effective retrieval strategy employing multi-resolution BOF to accelerate the match. The main idea is to improve the overall retrieval efficiency of BOF Descriptive Vector via the construction of BOF Low-resolution Vector and the comparison under low resolution to filter high-resolution candidate vectors. Based on stratified construction, we have improved uniform quantization multi-resolution BOF and proposed a non-linear Non-uniform quantization multi-resolution BOF method, which is combined with VA-file. At last, K-nearest neighbor retrieval algorithm is given. Experiments prove that this method has effectively increased the retrieval efficiency, improved the I/O function when loading mass image datasets and raised the system efficiency.
Digital Object Identifier (DOI)
Lipin, Wang and Juncheng, Pu
"Image Retrieval based on VA-File and Multi-Resolution BOW,"
Applied Mathematics & Information Sciences: Vol. 09:
1, Article 59.
Available at: https://digitalcommons.aaru.edu.jo/amis/vol09/iss1/59