Efficient and Effective Total Variation Image Super-Resolution: A Preconditioned Operator Splitting Approach

Table 1

The PSNR, ReErr, number of iterations, and computational times of the reconstructed images using three methods. The numbers in the bracket in the “Iterations” row refer to the total number of inner conjugate gradient iterations.

Experiments		Figure 2(a)	Figure 3(a)	Figure 4(a)	Figure 5(a)	Figure 6(a)	Figure 7(a)	Figure 8(a)
LDFP method	PSNR (dB)	27.4137	27.1968	25.7712	25.6881	24.8985	26.3618	21.9450
	ReErr	0.0056	0.0059	0.0094	0.0096	0.0115	0.0071	0.0164
	Iterations	46 (20)	44 (20)	49 (20)	49 (20)	43 (20)	49 (20)	44 (20)
	Time(s)	25.5156	24.5313	27.3125	27.2969	30.1250	34.2031	25.1406
AMM method	PSNR (dB)	28.5037	28.0035	26.3508	26.1247	24.3680	26.3947	22.0199
	ReErr	0.0044	0.0049	0.0082	0.0087	0.0130	0.0071	0.0161
	Iterations	189 (7)	176 (7)	239 (7)	228 (7)	209 (7)	184 (7)	131 (7)
	Time(s)	20.3906	18.9531	25.8750	24.6250	79.0625	69.5938	14.4531
OSM method	PSNR	28.8820	28.2968	26.5640	26.3086	24.8782	26.5322	22.0065
	ReErr	0.0040	0.0046	0.0079	0.0083	0.0116	0.0069	0.0162
	Iterations	90	93	100	101	194	183	58
	Time(s)	5.2031	5.3438	5.7656	5.8438	18.9531	17.8281	3.3750