Fig. 1. Structure of spatial heterodyne spectrometer

Fig. 2. Deep convolutional neural network architecture and denoising process

Fig. 3. Continuous-wave spatial heterodyne interferograms with different degrees of Gaussian noise

Fig. 4. The spectra of continuous-wave spatial heterodyne interferograms with different degrees of Gaussian noise

Fig. 5. The denoising results ((a)~(e)) and evaluate (f) of SHI-DnCNN on different Gaussian noise interferograms

Fig. 6. The spectra of interferograms after SHI-DnCNN denoising ((a)~(e)) and its spectral difference (f)

Fig. 7. Denoised interferogram of different algorithms for Sigma=25 Gaussian noise ((a)~(e)) and evaluate (f)

Fig. 8. The spectra of interferograms after different algorithms denoising ((a)~(e)) and its spectral difference (f)

Fig. 9. The results of denoising polychromatic spatial heterodyne interferograms with Sigma=25 Gaussian noise using SHI-DnCNN with different layers. (a) 12 layers; (b) 13 layers; (c) 14 layers; (d) 15 layers; (e) 16 layers; (f) 17 layers; (g) 18 layers; (h) 19 layers; (i) 20 layers; (j) 21 layers; (k) 22 layers; (l) 23 layers

Fig. 10. The results of denoising polychromatic spatial heterodyne interferograms with Sigma=50 Gaussian noise using SHI-DnCNN with different layers

Fig. 11. Comparison of training time (a), PSNR (b), SSIM (c), and spectral difference (d) of SHI-DnCNN with different layers

Fig. 12. (a) The CO₂ spatial heterodyne interferogram of GMI; (b) CO₂ spectral data

Fig. 13. The CO₂ spatial heterodyne interferograms with different degrees of Gaussian noise ((a1)~(c1)) and its spectrum ((a2)~(c2))

Fig. 14. Denoised results of CO₂ spatial heterodyne interferograms with different levels of Gaussian noise by SHI-DnCNN