Dynamic Path-Controllable Deep Unfolding Network for Compressive Sensing

Deep unfolding network (DUN) that unfolds the optimization algorithm into a deep neural network has achieved great success in compressive sensing (CS) due to its good interpretability and high performance. Each stage in DUN corresponds to one iteration in optimization. At the test time, all the sampling images generally need to be processed by all stages, which comes at a price of computation burden and is also unnecessary for the images whose contents are easier to restore. In this paper, we focus on CS reconstruction and propose a novel Dynamic Path-Controllable Deep Unfolding Network (DPC-DUN). DPC-DUN with our designed path-controllable selector can dynamically select a rapid and appropriate route for each image and is slimmable by regulating different performance-complexity tradeoffs. Extensive experiments show that our DPC-DUN is highly flexible and can provide excellent performance and dynamic adjustment to get a suitable tradeoff, thus addressing the main requirements to become appealing in practice. Codes are available at https://github.com/songjiechong/DPC-DUN.

I. Introduction

Compressive sensing (CS) is a considerable research interest from signal/image processing communities as a joint acquisition and reconstruction approach [1], [2]. The signal is first sampled and compressed simultaneously with random linear transformations. Then, the original signal can be reconstructed from far fewer measurements than that required by Nyquist sampling rate [3], [4], [5]. As CS can reduce the amount of information to be observed and processed while maintaining a reasonable reconstruction of the sparse or compressible signal, it has spawned many applications, including but not limited to medical imaging [6], [7], image compression [8], single-pixel cameras [9], [10], remote sensing [11], image classification [12], and snapshot compressive imaging [13], [14].