. 2019 Jul 18;7:e7075. doi: 10.7717/peerj.7075

Table 1. Set of features used by FSMKL, organized by importance.

The identified features were: the position (POS), the accumulated weight and the numbers of kernels in which they were used. All the features were identified using the terminology proposed in Machado et al. (2015).

Pos	Feature	Weight	Kernels	Pos	Feature	Weight	Kernels
1	Fractal(NoFilter(S),High)^a	2.627	3	12	JPEG(Canny(S),High)^a	0.747	4
2	Fractal(NoFilter(V),High)	2.627	3	13	JPEG(Canny(S),Medium)	0.747	4
3	Fractal(NoFilter(V),Medium)	2.216	2	14	JPEG(NoFilter(S),Medium)	0.483	2
4	Rank(NoFilter(S),R2)^a	2.122	4	15	Fractal(Canny(S),High)^a	0.393	2
5	Rank(NoFilter(S),M)	2.122	4	16	Size(Canny(S),M)^a	0.358	4
6	JPEG(NoFilter(S),High)^a	1.335	5	17	Size(Canny(V),M)	0.358	4
7	JPEG(NoFilter(V),High)	1.335	3	18	JPEG(Canny(S),Low)	0.326	4
8	Size(NoFilter(S),M)	1.195	4	19	Size(NoFilter(V),M)	0.149	2
9	Size(NoFilter(H+CS),R2)^a	1.195	4	20	Size(NoFilter(V),R2)	0.149	2
10	Fractal(Canny(S),Low)	0.883	3	21	Rank(Canny(S),R2)	0.128	4
11	Fractal(Canny(S),Medium)	0.883	3	22	Rank(Canny(S),M)^a	0.128	1

Notes.

Features previously identified by Machado et al. (2015) as the best individual features for solving the problem.