Scalable Deep Compressive Sensing

• DOI: 10.48550/arxiv.2101.08024
• Publication Date: 2021-01-01
• Authors: Zhonghao Zhang; Yipeng Liu; Xingyu Cao; Fei Wen; Ce Zhu
• Keywords: Initialization; Matrix (chemical analysis)

Deep learning has been used to image compressive sensing (CS) for enhanced reconstruction performance. However, most existing deep methods train different models subsampling ratios, which brings additional hardware burden. In this paper, we develop a general framework named scalable (SDCS) the sampling and (SSR) of all end-to-end-trained models. proposed way, images are measured initialized linearly. Two masks introduced flexibly control ratios in reconstruction, respectively. To make model adapt any ratio, training strategy dubbed is developed. training, trained with matrix initialization at various by integrating masks. Experimental results show that SDCS can achieve SSR without changing their structure while maintaining good performance, outperforms other methods.