In the tunnel magnetoresistance (TMR)-based electromagnetic tomography (EMT) (TMR-EMT) system, TMR measurements of alternating magnetic induction around the object field can reconstruct the permeability distribution image within the object. However, TMR’s nonlinear characteristics affect the accuracy of magnetic induction measurements and image reconstruction quality. Existing solutions to TMR nonlinearity often involve complex sensor circuit designs or extensive data fitting, making them cumbersome to implement. To address this issue, this article proposes a method to compensate for magnetic induction measurements using the information on odd harmonic components in the TMR output signal. The method is simple and easy to implement as it obtains the quantitative relationship between the amplitudes of the odd harmonic components and the magnetic induction measurements by formula derivation and is realized based on field-programmable gate array (FPGA) and host computer. Simulation results show that this method reduces the mean relative error of TMR measurements for an alternating magnetic induction amplitude range of 0.1–3 mT from 38% to 4.2% and decreases the nonlinear error from 70% to 3.3%. Experimental results indicate that the average correlation coefficient (CC) of the image reconstruction results for five magnetic permeability distributions improves from 0.8139 to 0.8495 after compensating for TMR-EMT’s nonlinear measurements.