DTCTH: a discriminative local pattern descriptor for image classification

EURASIP Journal on Image and Video Processing

Table 2 Effect of different SVM and DTCTH parameters on UIUC Sports Event, Caltech 101, and Scene 15 datasets

Techniques	UIUC sports event	Caltech 101	Scene 15
Linear kernel
DTCTH _8,1	85.16 ±0.96	72.26 ±1.67	82.66 ±0.50
D T C T H _8,2	84.73 ±1.01	76.08 ±0.41	82.87 ±0.49
Polynomial kernel
D T C T H _8,1	83.69 ±0.97	68.64 ±0.53	80.92 ±0.12
D T C T H _8,2	84.02 ±1.15	73.21 ±0.58	82.62 ±0.56
RBF kernel
D T C T H _8,1	75.74 ±1.54	58.72 ±1.16	72.95 ±0.62
D T C T H _8,2	75.83 ±1.17	63.96 ±1.36	73.73 ±0.65
Sigmoid kernel
D T C T H _8,1	67.95 ±1.94	52.59 ±1.41	68.16 ±1.88
D T C T H _8,2	70.47 ±1.58	56.88 ±1.31	69.59 ±1.08
Histogram intersection kernel
D T C T H _8,1	88.18 ±0.84	78.56 ±0.91	83.63 ±0.21
D T C T H _8,2	87.75 ±0.57	80.36 ±0.24	83.92 ±0.43

Here, we consider 70 training and 60 test images for UIUC Sports Event, 30 training and remaining test images for Caltech 101, and 100 training and remaining test images for Scene 15