Fig. 1. (a) Proof-of-principle schematic of complementary single-pixel imaging against the source’s energy fluctuation and (b) a previous method used for discussion.

Fig. 2. Different experimental detection signals in the condition of ε = 26.5 dB and δ = 0.24. (a) The target’s ideal detection signal without noise Y_t, (b) the signal Y_up detected by the detector D_up, (c) the signal Y_CSPI obtained by CSPI method, (d) the signal $Y_{{\mathrm{CSPI}}_{\mathrm{correction}}}=\frac{Y_{\mathrm{CSPI}}^i}{I_c^i}$ achieved by CSPI_correction method, and (e) the signal $Y_{{\mathrm{SPI}}_{\mathrm{correction}}}=\frac{Y^i}{I_m^i}$ achieved by SPI_correction method.

Fig. 3. Experimental demonstration of the influence of the source’s energy fluctuation δ on different reconstruction SPI methods when the DSNR ε is 26.5 dB. (a) δ = 0.03, (b) δ = 0.07, (c) δ = 0.11, (d) δ = 0.16, and (e) δ = 0.24. (f) The curve of PSNR-δ, where SPI_up is the reconstruction result based on Aⁱ(x) and $Y_{\text{up}}^i$, which corresponds to the conventional SPI.

Fig. 4. Effect of DSNR ε on the results of CSPI_correcion and SPI_correcion when δ = 0.2 is fixed. (a) ε = 10 dB, (b) ε = 15 dB, (c) ε = 20 dB, (d) ε = 25 dB, and (e) ε = 30 dB. (f) The curve of PSNR-ε.

Fig. 5. Performance comparison of different correction reconstruction methods in the condition of δ = 0.2. (a) SPI_correction with ε = 20 dB, (b) SPI_correction with ε = 21.5 dB, (c) CSPI_correction with ε = 20 dB, (d) SPI_up with ε = 20 dB, and (e) SPI_{up-correction} with ε = 20 dB, where SPI_{up-correction} is the reconstruction result based on Aⁱ(x) and $\frac{Y_{\text{up}}^i}{I_c^i}$.

Fig. 6. Simulated demonstration of imaging a complicated image “Lena” at different δ when the DSNR ε is 28 dB. The description of (a)–(f) is the same as Fig. 3.