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

Table 1

Mean PSNR (dB) and SSIM comparisons on Set11 at different CS ratios [16].

Algorithm	CR = 1%		CR = 4%		CR = 10%		CR = 25%
	PSNR	SSIM	PSNR	SSIM	PSNR	SSIM	PSNR	SSIM
STP-Net [25]	20.73	0.5111	23.39	0.6310	26.08	0.7679	30.06	0.8843
DPA-Net [22]	18.05	0.5011	23.50	0.7205	26.99	0.8354	31.74	0.9238
SWDGAN [23]	21.01	0.5410	24.67	0.6410	28.45	0.8570	33.45	0.9300
PE-Net [24]	20.74	0.4909	25.19	0.7586	28.58	0.8701	33.23	0.9407
ISTA-Net^+ [26]	17.42	0.4029	21.32	0.6037	26.64	0.8087	32.59	0.9254
OPINE-Net [26]	19.87	0.5070	25.04	0.7730	29.33	0.8825	34.44	0.9491
AMP-Net [30]	20.20	0.5581	25.26	0.7722	29.40	0.8779	34.63	0.9481
STP-ISTA-Net [16]	21.32	0.5529	25.47	0.7152	29.01	0.8438	33.72	0.9327
STP-DEQ-Net [16]	21.24	0.5565	24.67	0.7028	28.96	0.8495	32.63	0.9331
MsDC-DEQ-Net	21.43	0.5642	25.55	0.7303	29.96	0.8723	34.09	0.9446

The best performance is labeled in bold.