Fig. 3: Quantitative evaluation under different levels of composite degradation, including multiplicative illumination noise, spatial blur (isotropic and anisotropic), random jitter, and additive detection noise (Poisson and Gaussian). | Light: Science & Applications

Fig. 3: Quantitative evaluation under different levels of composite degradation, including multiplicative illumination noise, spatial blur (isotropic and anisotropic), random jitter, and additive detection noise (Poisson and Gaussian).

From: Comprehensive compensation of real-world degradations for robust single-pixel imaging

Fig. 3

The comparison includes PCM-DRGI33, DDPMGI32, GAN-SRSPI34, and the proposed BSRSPI. Across all metrics, the proposed BSRSPI consistently outperforms competing methods, demonstrating superior robustness in real-world scenarios. Arrows (↑) indicate that a higher value is better

Back to article page