EIT Image Reconstruction Method Based on DnCNN

The inverse problem of Electrical Impedance Tomography (EIT) is a highly ill-posed nonlinear problem. In order to obtain reconstructed images with good edge preservation, a DnCNN deep imaging method is proposed, which consists of a pre-reconstruction step and a denoising convolutional neural network (DnCNN) block. Tikhonov method is used to obtain a rough reconstruction. A single residual unit is used in DnCNN to predict the noise mapping in the pre-reconstruction result and recover a high-quality reconstructed image. The reconstruction results show that the DnCNN method proposed can effectively remove artifacts in the reconstructed image and accurately recover the position and edge of multiple target objects contained in the field. For the robust samples and the noise-added samples, the proposed method exhibits excellent anti-noise ability and generalization ability.