Multilevel wavelet-based hierarchical networks for image compressed sensing

作者：

Highlights：

• A hierarchical deep CS network based on multilevel wavelet transform is proposed.

• The sampling module is designed to adaptively acquire compressed measurements from a sparse signal.

• An interesting discovery in the use of the multilevel wavelet transform is that the structural sparsity can effectively accelerate the joint learning of the sampling network and reconstruction network for CS.

• By enhancing the reconstruction of finer texture details, our method achieves better performance.

摘要

•A hierarchical deep CS network based on multilevel wavelet transform is proposed.•The sampling module is designed to adaptively acquire compressed measurements from a sparse signal.•An interesting discovery in the use of the multilevel wavelet transform is that the structural sparsity can effectively accelerate the joint learning of the sampling network and reconstruction network for CS.•By enhancing the reconstruction of finer texture details, our method achieves better performance.

论文关键词：Compressed sensing,Hierarchical reconstruction,Sparse signal,Multilevel wavelet transform

论文评审过程：Received 11 December 2021, Revised 11 April 2022, Accepted 27 April 2022, Available online 29 April 2022, Version of Record 10 May 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108758