Table 1.

Segmentation results on the test set ^*

Method	F-measure	Kappa	Accuracy	Az
The proposed method	0.5228 (± 0.07)	0.4987 (± 0.07)	0.9337 (± 0.05)	0.8647 (± 0.06)
Matched filters b	0.489 (± 0.09)	0.4646 (± 0.09)	0.9322 (± 0.05)	0.7977 (± 0.08)
Local entropy b,†	0.4504 (± 0.11)	0.4049 (± 0.16)	0.8839 (± 0.19)	0.7104 (± 0.1)
Matched filters a,†	0.3847 (± 0.17)	0.3313 (± 0.2)	0.7481 (± 0.34)	0.7682 (± 0.1)
GMM Gabor b,†	0.3234 (± 0.19)	0.3046 (± 0.19)	0.9341 (± 0.04)	0.7921 (± 0.17)
GMM Gabor a,†	0.2861 (± 0.2)	0.2652 (± 0.19)	0.9304 (± 0.04)	0.7716 (± 0.18)
Local entropy a,†	0.2808 (± 0.23)	0.2545 (± 0.22)	0.892 (± 0.17)	0.7106 (± 0.1)
K-means Gabor b,†	0.1777 (± 0.12)	0.1667 (± 0.12)	0.9328 (± 0.04)	0.7727 (± 0.16)
K-means Gabor a,†	0.1536 (± 0.12)	0.1411 (± 0.12)	0.9308 (± 0.04)	0.7599 (± 0.17)

^*The results include the twenty-two test images: eleven manual frames and eleven automatic mosaics. Existing methods were applied to both the raw frames and the frames pre-processed with DLCF.

^aRaw frames

^bPre-processed frames

^†F-measure: p < 0.05