MsDC-DEQ-Net: Deep Equilibrium Model (DEQ) with Multiscale Dilated Convolution for Image Compressive Sensing (CS)

Table 2

Mean PSNR (dB) and SSIM comparisons on BSD68 at different CS ratios.

Algorithm	CR = 1%		CR = 4%		CR = 10%		CR = 25%
	PSNR	SSIM	PSNR	SSIM	PSNR	SSIM	PSNR	SSIM
STP-Net [25]	21.06	0.4639	23.11	0.5804	25.11	0.7091	28.48	0.8473
DPA-Net [22]	18.98	0.4643	23.27	0.6096	25.57	0.7267	29.68	0.8763
SWDGAN [23]	22.28	0.5600	24.88	0.6700	27.29	0.7890	30.96	0.9100
PENet [24]	22.10	0.4820	25.06	0.6711	27.38	0.7960	30.97	0.9053
ISTA-Net^+ [26]	19.14	0.4158	22.17	0.5486	25.32	0.7022	29.36	0.8525
OPINE-Net [26]	21.80	0.4972	24.87	0.6709	27.54	0.7966	31.28	0.9034
AMP-Net [30]	22.28	0.5387	25.26	0.6760	27.86	0.7926	31.74	0.9048
STP-ISTA-Net [16]	21.87	0.5046	25.12	0.6639	27.94	0.7932	31.35	0.8956
STP-DEQ-Net [16]	21.38	0.4882	23.74	0.6209	26.58	0.7615	30.48	0.8966
MsDC-DEQ-Net	22.53	0.5316	25.18	0.6724	28.24	0.7988	32.17	0.9136

The best performance is labeled in bold.