Robust Detection Model of Vascular Landmarks for Retinal Image Registration: A Two-Stage Convolutional Neural Network

Table 5

Comparison of the fundus image registration performance on FIRE database.

Method	Classes of FIRE data
	S	P	A
SIFT [13]	0.967	0.411	0.534
SURF [50]	0.978	0.384	0.422
GDB-ICP [51]	0.814	0.303	0.303
Harris-PIIFD [3]	0.900	0.090	0.443
H-M 16 [52]	0.945	0.443	0.577
REMPE [53]	0.958	0.542	0.660
Proposed method	0.988	0.511	0.803