Secondary image reconstruction based on Associative Markov Networks for electrical resistance tomography

The images reconstructed by electrical resistance tomography for two-phase flow with distinctive phase origins are usually blurred at the phase interface. To improve the image quality, secondary image reconstruction with Associative Markov Networks (AMNs) is presented. The initial images are reconstructed by the Landweber iteration algorithm. The obtained images are then processed using AMNs. The weights of AMNs are learned by a quadratic program and then a min-cut is used for the maximum a posteriori inference to obtain the optimal images. Simulation results from both noise-free and noisy data show significant improvement in the phase interface of images. For some conductivity distributions, the image errors can be reduced to a fifth of the initial values.