Robust Homography Estimation Based on Nonlinear Least Squares Optimization

Table 1

Homography estimation results on random generated false matches. TP and FP mean true positive and false positive number, while TN and FN represent true negative and false negative number. The column “RMSE w/o S ^” represents the RMSE of approach without using descriptor similarity S, while “RMSE w/o C ^” represents the RMSE of standard optimization approach whose cost function is introduced in (6).

True matches	False matches	TP	FP	TN	FN	RMSE	RMSE w/o S	RMSE w/o C
42	0	42	0	0	0	0.166	0.165	0.184
42	51	42	1	50	0	0.573	0.462	13.992
42	103	42	1	102	0	0.597	0.699	21.420
42	154	42	1	153	0	0.222	0.608	27.589
42	206	42	1	205	0	0.633	5.816	30.845
42	257	42	1	256	0	0.622	17.311	34.856
42	309	42	1	308	0	0.228	28.479	37.672
42	360	41	1	359	1	0.652	36.286	38.408
42	412	42	1	411	0	0.594	37.278	40.348
42	463	41	1	462	1	0.821	39.889	41.875
42	515	42	1	514	0	0.825	38.737	42.467