Back to Algorithm Ranking.

[Grayscale] Boundary Detection Benchmark: Algorithm "Second Moment Matrix"

This algorithm finds boundaries by examining the eigenspectrum of the spatially averaged second moment matrix. Image gradients are estimated with a pair of Gaussian derivative filters. The two eigenvalues are used as features in a logistic model to estimate the posterior probability of a boundary at each pixel. For these 480x320 pixel images, the optimal scale is sigma=2 (for both inner and outer scales). If we order the eigenvalues from smaller to larger, the logistic coefficients are -7.33 and 30.9, with an offset of -3.25.

Click on an image for additional details.

#1 (119082) F=0.68
#2 (170057) F=0.63
#3 (58060) F=0.41
#4 (163085) F=0.43
#5 (42049) F=0.89
#6 (167062) F=0.76
#7 (157055) F=0.70
#8 (295087) F=0.56
#9 (24077) F=0.70
#10 (78004) F=0.75
#11 (220075) F=0.59
#12 (45096) F=0.71
#13 (38092) F=0.69
#14 (43074) F=0.61
#15 (16077) F=0.49
#16 (86000) F=0.64
#17 (101085) F=0.74
#18 (219090) F=0.64
#19 (89072) F=0.65
#20 (300091) F=0.76
#21 (126007) F=0.65
#22 (156065) F=0.54
#23 (76053) F=0.61
#24 (296007) F=0.57
#25 (175032) F=0.43
#26 (253027) F=0.63
#27 (304034) F=0.38
#28 (86016) F=0.48
#29 (103070) F=0.55
#30 (8023) F=0.30
#31 (260058) F=0.53
#32 (41033) F=0.60
#33 (291000) F=0.52
#34 (109053) F=0.54
#35 (130026) F=0.52
#36 (241004) F=0.75
#37 (108082) F=0.35
#38 (285079) F=0.68
#39 (147091) F=0.70
#40 (69040) F=0.47
#41 (14037) F=0.56
#42 (54082) F=0.54
#43 (189080) F=0.77
#44 (229036) F=0.60
#45 (62096) F=0.77
#46 (271035) F=0.68
#47 (167083) F=0.43
#48 (12084) F=0.48
#49 (69015) F=0.74
#50 (148089) F=0.67
#51 (160068) F=0.66
#52 (145086) F=0.64
#53 (216081) F=0.72
#54 (97033) F=0.64
#55 (182053) F=0.72
#56 (208001) F=0.61
#57 (19021) F=0.62
#58 (227092) F=0.75
#59 (134035) F=0.69
#60 (223061) F=0.39
#61 (253055) F=0.53
#62 (148026) F=0.55
#63 (210088) F=0.69
#64 (86068) F=0.57
#65 (3096) F=0.86
#66 (41069) F=0.61
#67 (21077) F=0.67
#68 (196073) F=0.67
#69 (108070) F=0.33
#70 (123074) F=0.47
#71 (376043) F=0.51
#72 (306005) F=0.63
#73 (38082) F=0.58
#74 (33039) F=0.53
#75 (108005) F=0.44
#76 (106024) F=0.67
#77 (302008) F=0.57
#78 (102061) F=0.53
#79 (197017) F=0.77
#80 (299086) F=0.67
#81 (37073) F=0.78
#82 (241048) F=0.65
#83 (65033) F=0.58
#84 (55073) F=0.54
#85 (66053) F=0.67
#86 (143090) F=0.66
#87 (85048) F=0.70
#88 (42012) F=0.52
#89 (351093) F=0.66
#90 (361010) F=0.68
#91 (175043) F=0.29
#92 (87046) F=0.54
#93 (105025) F=0.46
#94 (236037) F=0.30
#95 (101087) F=0.70
#96 (304074) F=0.59
#97 (296059) F=0.86
#98 (159008) F=0.53
#99 (385039) F=0.71
#100 (69020) F=0.64

Page generated on 20-Feb-2013 11:08:18.