AutoBCS: Block-based Image Compressive Sensing with Data-driven Acquisition and Non-iterative Reconstruction

Block compressive sensing is a well-known signal acquisition and reconstruction paradigm with widespread application prospect of science, engineering and cybernetic systems. However, the state-of-the-art block-based image compressive sensing (BCS) generally suffer from two issues. The sparsifying domain and the sensing matrices widely used for image acquisition are not data-driven, thus ignoring both the features of the image and the relationship among sub-block images. Moreover, it requires to address high-dimensional optimization problem with extensive computational complexity for image reconstruction. In this paper, we provide a deep learning strategy for BCS, called AutoBCS, which takes into account the prior knowledge of image in the acquisition step and establishes a subsequent reconstruction model to obtain fast image reconstruction with low computational cost. More precisely, we present a learning-based sensing matrix (LSM) from training data so as to accomplish image acquisition, therefore capturing and preserving more image characteristics. In particular, the generated LSM is proved to satisfy the theoretical requirements, such as the so-called restricted isometry property. Additionally, we build a non-iterative reconstruction network, which provides an end-to-end BCS reconstruction to eliminate blocking artifacts and maximize image reconstruction accuracy, in our AutoBCS architecture. Furthermore, we investigate comprehensive comparison studies with both traditional BCS approaches as well as newly-developing deep learning methods. Compared with these approaches, our AutoBCS framework can not only provide superior performance in both image quality metrics (SSIM and PSNR) and visual perception, but also automatically benefit reconstruction speed.