Extended Spectral Regression for efficient scene recognition
作者：
Highlights：
• A new Spectral Regression approach, Extended Spectral Regression (ESR), for SR-based manifold learning on a large dataset.
• A new approach to compute the edge weights for graph learning on tiny clusters containing samples of various classes.
• A novel method based on ESR for efficient scene recognition.
• An enhanced low-level feature representation containing various aspects of local features.
• An ESR learning algorithm to effectively embed low-level image features for BOW-based image representation.
摘要
Highlights•A new Spectral Regression approach, Extended Spectral Regression (ESR), for SR-based manifold learning on a large dataset.•A new approach to compute the edge weights for graph learning on tiny clusters containing samples of various classes.•A novel method based on ESR for efficient scene recognition.•An enhanced low-level feature representation containing various aspects of local features.•An ESR learning algorithm to effectively embed low-level image features for BOW-based image representation.
论文关键词：Spectral Regression,Subspace learning,Machine learning,Image classification,Scene recognition,Computer vision
论文评审过程：Received 5 October 2012, Revised 1 October 2013, Accepted 18 March 2014, Available online 27 March 2014.
论文官网地址：https://doi.org/10.1016/j.patcog.2014.03.012
