SAR Image Matching Based on Local Feature Detection and Description Using Convolutional Neural Network

Table 3

Performance results of second dataset.

Study case	Metric	Before	CC	SIFT	ORB	Proposed	[12]
Subset	Shift	−20, 21	−2, 7	0, 0	0, 0	0, 0	0, 0
	MSE	95.99	96.13	31.48	48.58	83.03	85.34
	PSNR	14.154	14.151	16.575	15.633	14.445	14.409
	CC	−0.012	0.122	0.721	0.708	0.433	0.398
Translation	Shift	5, −23	0, −20	0, 0	0, 0	0, 0	0, 0
	MSE	95.35	95.55	58.42	63.27	82.52	86.45
	PSNR	14.168	14.164	15.233	15.059	14.483	14.381
	CC	−0.023	−0.036	0.716	0.704	0.534	0.319
Scaling	Shift	0, 0	0, 0	0, 0	0, 0	0, 0	0, 0
	MSE	3.60	58.13	45.16	69.12	60.88	70.81
	PSNR	21.282	15.243	15.791	14.867	15.143	14.814
	CC	−0.023	0.963	0.981	0.841	0.939	0.881
Rotation	Shift	8, −21	−24, 0	0, 0	0, 0	0, 0	0, 0
	MSE	92.68	90.51	63.95	66.34	91.36	87.83
	PSNR	14.230	14.281	15.036	14.957	14.261	14.347
	CC	−0.101	−0.008	0.949	0.926	−0.141	0.171