Scene search based on the adapted triangular regions and soft clustering to improve the effectiveness of the visual-bag-of-words model

EURASIP Journal on Image and Video Processing

Table 1 MAP performance on the image benchmark of the Corel-1K with a step size = 10

Features % per image for dictionary formation	MAP performance analysis on different dictionary sizes
Features % per image for dictionary formation	20	60	100	200	400
10%	80.98	83.42	85.14	87.03	86.81
25%	81.08	83.58	85.21	87.14	86.68
50%	81.41	83.34	85.33	87.19	86.54
75%	81.78	83.78	85.22	87.22	86.64
100%	81.92	83.86	85.46	87.21	86.51
MAP	81.43	83.59	85.27	87.15	86.63
Std. dev.	0.41	0.22	0.12	0.16	0.11
Conf. interval	80.91–81.94	83.31–83.87	85.11–85.42	87.01–87.42	86.48–86.78
Std. error	0.18	0.10	0.05	0.07	0.05