Bag-of-Words Model for Image Classification Based on Harris Corner Features Weighting

The dense Scale-Invariant Feature Transform (SIFT) algorithm for Bag-of-Words (BOW) model may fail to extract high discriminating and representative features. To solve this problem, a new algorithm based on corner features weighting for BOW model image classification was proposed. The corner SIFT features were extracted in the light of each position, and certain weights for corner SIFT features were set on the basis of the corner degrees. This could not only distinguish the contribution of different corner features, but also make the corner SIFT features more prominent compared with the dense SIFT features. The experimental results showed that the BOW model constructed by the proposed approach worked well for image classification.