Joint learning sparsifying linear transformation for low-resolution image synthesis and recognition

作者：

Highlights：

• This paper presents a method that has the capability of solving two problems simultaneously, image super-resolution and classification.

• The sparse transformation matrix learned by our proposed method could capture task-specific discriminative information of images that is not easily accessible in the original images.

• The proposed learning model has been successfully applied to low-resolution image classification with both supervised and semi-supervised settings.

摘要

Highlights•This paper presents a method that has the capability of solving two problems simultaneously, image super-resolution and classification.•The sparse transformation matrix learned by our proposed method could capture task-specific discriminative information of images that is not easily accessible in the original images.•The proposed learning model has been successfully applied to low-resolution image classification with both supervised and semi-supervised settings.

论文关键词：Sparse representation,Joint dictionary learning,Sparse linear transformation,Geometric optimization,Low-resolution image classification

论文评审过程：Received 9 August 2016, Revised 30 November 2016, Accepted 9 January 2017, Available online 12 January 2017, Version of Record 12 March 2017.

论文官网地址：https://doi.org/10.1016/j.patcog.2017.01.013