A Novel Technique Based on Visual Words Fusion Analysis of Sparse Features for Effective Content-Based Image Retrieval
Table 1
Statistical analysis and MAP performance of the proposed technique based on visual words fusion on different dictionary sizes and features percentages per image (bold values indicate the best performance).
Features percentages per image
Performance analysis in terms of the MAP performance (in %) on the different sizes of the dictionary
20
50
100
200
400
600
800
1000
1200
10%
74.30
74.60
76.10
78.50
79.30
80.70
84.50
76.60
75.30
25%
75.30
75.60
76.20
79.20
79.60
81.60
84.60
77.40
75.60
50%
75.60
76.00
76.50
80.60
81.70
82.30
87.30
77.50
76.30
75%
75.80
76.30
77.50
81.30
82.10
83.01
85.60
78.20
76.40
100%
76.00
77.60
79.10
81.70
82.30
83.60
85.70
78.50
77.30
MAP
75.40
76.10
77.10
80.20
81.00
82.24
85.90
77.64
76.18
Std. error
0.29
0.48
0.56
0.61
0.64
0.51
0.50
0.33
0.34
Std. deviation
0.66
1.09
1.25
1.36
1.43
1.14
1.12
0.74
0.77
Conf. interval
74.50–76.20
74.60–77.30
75.50–78.60
78.50–81.90
79.20–82.70
80.80–83.60
84.10–86.90
76.70–78.50
75.20–77.10
Statistical analysis using nonparametric Wilcoxon matched-pairs signed-rank test
