Modeling multiple visual words assignment for bag-of-features based medical image retrieval

Abstract
In this paper, we investigate the bag-of-features based medical image retrieval methods, which represent an image as a collection of local features, such as image patch and key points with SIFT descriptor. To improve the bag-of-features method, we first model the assignment of local descriptor as contribution functions, and then propose a new multiple assignment strategy. By assuming the local feature can be reconstructed by its neighboring visual words in vocabulary, we solve the reconstruction weights as a QP problem and then use the solved weights as contribution functions, which results in a new assignment method called the QP assignment. We carry our experiments on ImageCLEFmed datasets. Experiments' results show that our proposed method exceeds the performances of traditional solutions and works well for the bag-of-features based medical image retrieval tasks.

Citation
Wang, J., & Almasri, I. (2012). Modeling Multiple Visual Words Assignment for Bag-of-Features based Medical Image Retrieval. Signal Processing, Pattern Recognition and Applications / 779: Computer Graphics and Imaging. doi:10.2316/p.2012.779-015

Publisher
ACTA Press

Journal
Signal Processing, Pattern Recognition and Applications / 779: Computer Graphics and Imaging

Conference/Event Name
IASTED International Conference on Computer Graphics and Imaging, CGIM 2012

DOI
10.2316/P.2012.779-015

Permanent link to this record