Skip to main content

Table 1 Performance evaluation results of our proposed method against the TOP 3 winners of the annual DIBCO or H-DIBCO competitions (best results highlighted in bold)

From: An enhanced binarization framework for degraded historical document images

Dataset Method FM (%) pFM (%) PSNR (dB) NRM (%) DRD MPM (‰)
DIBCO
2009
Rank 1st [21] 91.24   18.66 4.31   0.55
Rank 2nd [55] 90.06   18.23 4.75   0.89
Rank 3rd [56] 89.34   17.79 5.32   1.90
Proposed 93.46   20.01 2.59   1.54
H-DIBCO
2010
Joint 1st [23] 91.50 93.58 19.78 5.98   0.49
Joint 1st [57] 89.70 95.15 19.15 8.18   0.29
Rank 2nd [58] 91.78 94.43 19.67 4.77   1.33
Rank 3rd [28] 89.73 90.11 18.90 5.78   0.41
Proposed 93.73 95.18 20.97 3.64   0.29
DIBCO
2011
Rank 1st [59] 80.86   16.13   104.48 64.43
Rank 2nd [24] 85.20   17.16   15.66 9.07
Rank 3rd [28] 88.74   17.84   5.36 8.68
Proposed 90.72   18.85   4.47 7.87
H-DIBCO
2012
Rank 1st [29] 89.47 90.18 21.80   3.44  
Rank 2nd [59] 92.85 93.34 20.57   2.66  
Rank 3rd [24] 91.54 93.30 20.14   3.05  
Proposed 94.26 95.16 21.68   2.08  
DIBCO
2013
Rank 1st [24] 92.12 94.19 20.68   3.10  
Rank 2nd [29] 92.70 93.19 21.29   3.18  
Rank 3rd [60] 91.81 92.67 20.68   4.02  
Proposed 93.51 94.54 21.32   2.77  
H-DIBCO
2014
Rank 1st [33] 96.88 97.65 22.66   0.90  
Rank 2nd [29] 96.63 97.46 22.40   1.00  
Rank 3rd [61] 93.35 96.05 19.45   2.19  
Proposed 96.77 97.73 22.47   0.95  
H-DIBCO
2016
Rank 1st [34] 87.61 91.28 18.11   5.21  
Rank 2nd [62, 63] 88.72 91.84 18.45   3.86  
Rank 3rd [62] 88.47 91.71 18.29   3.93  
Proposed 89.64 93.56 18.69   4.03  
DIBCO
2017
Rank 1st [45] 91.04 92.86 18.28   3.40  
Rank 2nd 89.67 91.03 17.58   4.35  
Rank 3rd [42] 89.42 91.52 17.61   3.56  
Proposed 89.37 90.80 17.99   5.51  
H-DIBCO
2018
Rank 1st [64] 88.34 90.24 19.11   4.92  
Rank 2nd 73.45 75.94 14.62   26.24  
Rank 3rd 70.01 74.68 13.58   17.45  
Proposed 88.34 90.37 19.11   4.93