Diffusion-Steered Algorithm for Improving Quality of Images Generated by Electrical Capacitance Tomography Measurement System

Abstract

Electrical Capacitance Tomography (ECT) is a non-invasive, non-intrusive, radiation-free, robust, and cost-efficient measurement system that generates cross-sectional images of industrial operations by measuring differences in dielectric properties within a container. Despite its advantages, the poor quality of reconstructed images limits its applicability. Researchers have explored both non-iterative and iterative methods to address this issue, yet the resulting image quality remains insufficient for high-stakes applications where accurate decision-making is critical. This study proposes a novel algorithm that integrates the ECT model with a diffusion-steered image denoising functional, enhancing the quality of ECT reconstructed images beyond existing methods. Empirical comparisons show that the proposed algorithm outperforms state-of-the-art techniques such as Linear Back Projection and Projected Landweber, with Distribution Error (DE) and Correlation Coefficient (CC) improvements of 62% and 19%, respectively. Qualitative assessments further indicate the superior performance of the proposed algorithm in reconstructing high-quality images.