No abstract

A well-established reconstruction algorithm for electrical impedance tomography uses a finite-element method to model the forwards problem using Neumann boundary conditions. The reconstruction is then obtained by solving the inverse problem through an iterative non-linear least-squares fit of the model electrode voltages to the measured voltages. It is also usual to apply Tikhonov regularization to improve the condition of the inverse problem. However, such regularization introduces artefacts into the solution estimate. We present a quasi-single-step reconstruction technique based on a weakly regularized solution constructed from the final Jacobian of the standard iterative scheme, with filtering to act as a `mollifier'. This reconstruction has well-defined spectral properties, since it approximates a filtered version of the original impedance distribution. Some illustrative results are given for 2D resistance tomography.

With the power of deep learning, super-resolution (SR) methods enjoy a dramatic boost in performance. However, they usually have a large model size and high computational complexity, which hinders the application in devices with limited memory and computing power. Some lightweight SR methods solve this issue by directly designing shallower architectures, but it will adversely affect the representation capability of convolutional neural networks. To address this issue, we propose the dual feature aggregation strategy for image SR. It enhances feature utilization via feature reuse, which largely improves the representation ability while only introducing marginal computational cost. Thus, a smaller model could achieve better cost-effectiveness with the dual feature aggregation strategy. Specifically, it consists of Local Aggregation Module (LAM) and Global Aggregation Module (GAM). LAM and GAM work together to further fuse hierarchical features adaptively along the channel and spatial dimensions. In addition, we propose a compact basic building block to compress the model size and extract hierarchical features in a more efficient way. Extensive experiments suggest that the proposed network performs favorably against state-of-the-art SR methods in terms of visual quality, memory footprint, and computational complexity.