Multi-Scale Deep Compressive Sensing Network

Abstract

With joint learning of sampling and recovery, the deep learning-based compressive sensing (DCS) has shown significant improvement in performance and running time reduction. Its reconstructed image, however, losses highfrequency content especially at low subrates. This happens similarly in the multi-scale sampling scheme which also samples more low-frequency components. In this paper, we propose a multi-scale DCS convolutional neural network (MS-DCSNet) in which we convert image signal using multiple scale-based wavelet transform, then capture it through convolution block by block across scales. The initial reconstructed image is directly recovered from multi-scale measurements. Multi-scale wavelet convolution is utilized to enhance the final reconstruction quality. The network is able to learn both multi-scale sampling and multi-scale reconstruction, thus results in better reconstruction quality.

An Overview on Deep Learning Techniques for Video Compressive Sensing

“…Shi et al [39] and T.N. Canh et al [40] proposed CNN-based methods for 2D image reconstruction that split the reconstruction process into two stages. Firstly, the initial reconstruction which aims to recover the images from the patches.…”

“…Deep compressive sensing was extended to multi-scale schemes [40][41][42] utilizing image decomposition. In [41], a multiphase reconstruction process is proposed.…”

Tensor-Based Light Field Compressed Sensing and Epipolar Plane Images Reconstruction via Deep Learning

“…network (ISAT-Net) to CS reconstruction of natural images in a fast way. Canh et al [35] proposed a multi-scale deep CS convolutional neural network to recover the initial reconstructed image and enhance the final reconstruction quality. These methods make it possible to apply deep learning approach to CS in the LF.…”

“…Recently, the deep learning method has attracted much attention and several deep learning-based CS methods are proposed [32]- [35]. The deep learning network can effectively solve the problem of designing a proper measurement matrix and reconstructing images.…”

“…In this paper, we focus on the CS-based LF compression and use the convolutional neural networks (CNN) to reconstruct LF images. The CNNbased methods for LF images reconstruction mainly divides the reconstruction part into two stages [32], [35]. In the first stage, the images will be recovered from the patches, which is named the initial reconstruction.…”