Person re-identification based on saliency
Cailing Wang, Song Yuan Tang, Songhao Zhu, Xiao‐Yuan Jing · 2016
Two novel salient features for person re-identification are proposed in this paper. Salient feature in image always provides valuable information for pattern recognition, therefore we seek two kinds of salient features to represent pedestrians, a Speeded Up Robust Features (SURF) and a Principal Component Analysis-Scale Invariant Feature Transform (PCA-SIFT) feature are used to describe texture of person image respectively, then texture information and color histogram construct the feature space. In the feature space, salient features are detected according to the difference between person images. The effectiveness of our salient features in people re-identification is validated on the widely used GRID dataset, we get more accurate and efficiency results using SURF and PCA-SIFT based salient features.