TY - CONF
T1 - Image Reconstruction Algorithm for Electrical Impedance Tomography Using Group Sparsity
AU - Yang, Yunjie
AU - Jia, Jiabin
PY - 2016/9
Y1 - 2016/9
N2 - Electrical Impedance Tomography (EIT) is one of electrical tomography modalities for non-intrusive, high-temporal-resolution conductivity imaging. Attributing to its superior characteristics, EIT has found it extensive research and applications in both industrial and biomedical fields. Despite its high temporal resolution, improvement for the relatively low reconstructed image quality in practical situations is intensely desired and has become a research focus for long. In this paper, an image reconstruction algorithm based on group sparsity constraint is introduced. As an extension of standard sparsity concept, the idea of group sparsity has been come up with in recent years, which further takes advantage of the fact that in many cases the non-zero coefficients in the reconstructed signal or image tend to be clustered rather than randomly distributed. In standard sparsity prior, a constraint on non-zeros is applied for the whole target image. While in EIT, conductivity variance usually appears as groups. Therefore, clustered sparsity prior is expected to be more effective to promote image feature than the standard sparsity regularization in EIT imaging. Motivated by this idea, the image reconstruction framework based on group sparsity for EIT is proposed and grouping methods is presented. Numerical analysis is performed on small scale conductivity phantoms to validate the algorithm. Furthermore, static experiments by our developed EIT system is also carried out. Comparison between the proposed algorithm and conventional Landweber method verifies the performance of proposed algorithm.
AB - Electrical Impedance Tomography (EIT) is one of electrical tomography modalities for non-intrusive, high-temporal-resolution conductivity imaging. Attributing to its superior characteristics, EIT has found it extensive research and applications in both industrial and biomedical fields. Despite its high temporal resolution, improvement for the relatively low reconstructed image quality in practical situations is intensely desired and has become a research focus for long. In this paper, an image reconstruction algorithm based on group sparsity constraint is introduced. As an extension of standard sparsity concept, the idea of group sparsity has been come up with in recent years, which further takes advantage of the fact that in many cases the non-zero coefficients in the reconstructed signal or image tend to be clustered rather than randomly distributed. In standard sparsity prior, a constraint on non-zeros is applied for the whole target image. While in EIT, conductivity variance usually appears as groups. Therefore, clustered sparsity prior is expected to be more effective to promote image feature than the standard sparsity regularization in EIT imaging. Motivated by this idea, the image reconstruction framework based on group sparsity for EIT is proposed and grouping methods is presented. Numerical analysis is performed on small scale conductivity phantoms to validate the algorithm. Furthermore, static experiments by our developed EIT system is also carried out. Comparison between the proposed algorithm and conventional Landweber method verifies the performance of proposed algorithm.
M3 - Paper
T2 - 8th World Congress on Industrial Process Tomography
Y2 - 26 September 2016 through 29 September 2016
ER -