As a non-invasive measurement technique, electromagnetic tomography (EMT) system based on tunneling magneto resistance (TMR) can detect the distribution of medium in the gas–liquid–solid fluidized bed by the difference of permeability. The distribution of medium in the three-phase fluidized bed is not uniform, and the solid clusters and bubble clusters are formed in the background of the gas–liquid–solid mixture. Since the effect of the background in the pipe of the fluidized bed on the boundary measurement data of the TMR-EMT system is much greater, it is difficult to detect solid clusters and bubble clusters when the image is reconstructed directly using the boundary measurement data. In order to improve the detection of clusters by the TMR-EMT system, a method to weaken the effect of the background is proposed. The equivalent magnetic circuit model is used to estimate the background permeability. Then the magnetic dipole theory and demagnetization effect theory are utilized to establish a simple background compensation model based on the estimated background permeability. Using the compensation model, the effect of the background is weakened from the boundary measurement data and the cluster distribution information is highlighted. Simulation and experimental results demonstrate the effectiveness of the proposed method.