A person re-identification algorithm by exploiting region-based feature salience

• A region-based person re-identification algorithm is proposed to fully exploit the salience of features.
• Part-based feature extraction is proposed to adopt different features for different parts based on their characteristics.
• A salient color descriptor is proposed by considering color diversity between current region and surrounding regions.
• The proposed salient color descriptor can achieve a robust local color feature representation.
• Experimental results shows that our algorithm outperform the similar algorithms published in CVPR 2013, CVPR 2010, etc.

Due to the changes of the pose and illumination, the appearances of the person captured in surveillance may have obvious variation. Different parts of persons will possess different characteristics. Applying the same feature extraction and description to all parts without differentiating their characteristics will result in poor re-identification performances. Therefore, a person re-identification algorithm is proposed to fully exploit region-based feature salience. Firstly, each person is divided into the upper part and the lower part. Correspondingly, a part-based feature extraction algorithm is proposed to adopt different features for different parts. Moreover, the features of every part are separately represented to retain their salience. Secondly, in order to accurately represent the color feature, the salient color descriptor is proposed by considering the color diversity between current region and its surrounding regions. The experimental results demonstrate that the proposed algorithm can improve the accuracy of person re-identification compared with the state-of-the-art algorithms.