Fig. 1. Schematic diagram of mesh generation and forward projection matrix calculation method for axisymmetric object reconstruction

Fig. 2. Ground truth linear attenuation coefficient distribution of FTO, the comparison between the noisy optical path and the true value, and the comparison between the noisy transmittance and the true value; in Fig (b), the noise is independently and equally distributed; in Fig (c), the noise is difficult to see; however in Fig (d), the noise is clearly visible

Fig. 3. The mean and uncertainty of FTO linear attenuation coefficient reconstructed by Gibbs algorithm and RTOiG algorithm, with noise level of R_SN=40, N=10000, burn-in number is 5000

Fig. 4. The autocorrelation coefficient figure of calculated with the MCMC chain sampled by Gibbs algorithm, for N=10000, burn-in number of 5000N=10000，退火数5000，利用Gibbs算法抽样得到的MCMC链，计算得到关于的自相关系数图

Fig. 5. The autocorrelation coefficient figure of calculated with the MCMC chain sampled by RTOiG algorithm, for N=10000, burn-in number of 5000N=10000，退火数5000，RTOiG算法抽样得到的MCMC链，关于的自相关系数图

Fig. 6. The noisy optical path generated by simulation, and the mean and uncertainty of FTO linear attenuation coefficient reconstructed by Gibbs algorithm under different R_SN, for N=10000 and burn-in number of 5000

Fig. 7. The noisy transmittance generated by simulation (note that the y axis is logarithmic coordinate), and the mean and uncertainty of FTO linear attenuation coefficient reconstructed by RTOiG algorithm under different R_SN, for N=10000 and burn-in number of 5000

Fig. 8. Ground truth linear attenuation coefficient distribution of pcv-STO

Fig. 9. The noisy optical path generated by simulation, and the mean and uncertainty of pcv-STO linear attenuation coefficient reconstructed by Gibbs algorithm under different R_SN, for N=10000 and burn-in number of 5000

Fig. 10. The noisy transmittance generated by simulation (note that the y axis is logarithmic coordinate), and the mean and uncertainty of pcv-STO linear attenuation coefficient reconstructed by RTOiG algorithm under different R_SN, for N=10000 and burn-in number of 5000