RETRACTED: Image Inpainting via Context Discriminator and U-Net

Table 4

Comparison of inpainting effect between proposed method and other image inpainting methods.

Input	Methods	L1	L2	PSNR	SSIM	MOS
I1	Iizuka et al.	13.8718	1.0414	30.2190	89.1628	0.8288
	Yu et al.	14.7369	1.7698	32.1979	92.3744	0.8312
	Wang et al.	21.7793	1.3783	34.8191	93.1767	0.8345
	Proposed	13.8470	0.9576	35.8945	93.9849	0.9011
I2	Iizuka et al.	15.2252	2.1583	28.7412	86.3417	0.8124
	Yu et al.	18.5086	3.2251	29.8930	90.2347	0.8233
	Wang et al.	23.6323	2.7219	31.0758	91.9835	0.8451
	Proposed	15.1322	2.0549	31.6694	92.7252	0.9505
I3	Iizuka et al.	17.5519	4.2346	25.7562	82.5773	0.8032
	Yu et al.	22.5854	5.6448	27.0268	84.9860	0.8213
	Wang et al.	25.0629	4.4443	29.9880	88.1916	0.8312
	Proposed	18.2379	4.0767	30.0075	89.0469	0.9120
I4	Iizuka et al.	19.7029	5.3181	24.1443	81.0334	0.8012
	Yu et al.	21.5413	7.0420	25.5798	83.5718	0.8122
	Wang et al.	27.6290	6.3969	27.9264	85.4103	0.8244
	Proposed	20.1965	6.0601	27.6131	86.5701	0.9266

Note. The bold number in Table 4 is the best experimental results for each row.