Research On Super Resolution Enhancement Algorithm Based On Sparse Imaging

Electrical capacitance tomography(ECT)is an important computed tomography technology,which has great research value in non-contact medical diagnosis,multiphase flow imaging and other fields.It has the advantages of fast imaging speed and low equipment cost.However,due to the less plate arrangement of the sensor equipment,the imaging effect is poor and the resolution is low.Based on the electrical capacitance tomography(ECT)algorithm,this paper proposes a new reconstruction method based on image super-resolution,and achieved a series of research results:(1)The experimental data set uses high sampling rate image of human head simulation image as label data and low sampling rate image data as input data.In traditional image super-resolution training,the difference of image information between high-resolution tag image and low-resolution input image is little,because the low-resolution image is obtained from high-resolution image after downsampling.We improved the problem of the difference between images.It improves the problem that the accuracy of direct transfer learning super-resolution algorithm is not improved.(2)In this paper,a simple IDN module is used as the feature extraction module.This ensures the lightweight advantage of ECT,and at the same time improves the problem of gradient disappearance caused by the deep design of most super-resolution models.At the same time,it also indirectly alleviates the contradiction between the depth of network layers and over fitting.(3)In this paper,a new loss function is proposed to improve the recovery accuracy.The design goal is that in the process of image restoration,the model can't generate high-resolution image irregularly without the basis of input.At the same time,it ensures that the model only modifies the differences between the high and low sampling rates,and reduces the modifications to the places without differences.The experimental results show that the proposed method can obtain higher peak signal-to-noise ratio(PSNR)for electrical capacitance tomography(ECT)image reconstruction,and the value on the experimental data set is more than 34dB.