Noise Resilient Local Gradient Orientation for Content-Based Image Retrieval

Table 7

Category-wise performance comparison of NRLGO at Corel 1k with state-of-the-art descriptors.

Category	Performance	MTH [44]	MSD [43]	SED [46]	ENN [47]	CDH [45]	CMSD [12]	LeNet-F6 [48]	Proposed
African	Precision	69.17	83.33	82.50	85	77.50	86.66	80	88.25
	Recall	8.30	10.00	9.90	9.00	9.30	10.40	8.00	10.30
Beach	Precision	61.67	43.33	28.33	75	56.67	42.08	60	50.75
	Recall	7.40	5.20	3.40	7.00	6.80	5.05	6.5	5.15
Building	Precision	45.83	63.33	47.50	70	47.50	81.66	75	85.00
	Recall	5.50	7.60	5.70	6.00	5.70	9.80	8.00	9.95
Bus	Precision	68.33	76.67	73.33	75	71.67	81.66	80	83.70
	Recall	8.20	9.20	8.80	7.00	8.60	9.80	9.00	9.95
Dinosaur	Precision	100.00	100.00	90.00	100	100	100	100	100
	Recall	12.00	12.00	10.80	12	12.00	12.00	12.00	12.00
Elephant	Precision	70.83	65.00	55.00	75	62.50	72.08	85	100
	Recall	8.50	7.80	6.60	7.00	7.50	8.65	10.00	12.00
Flower	Precision	75.00	86.67	72.50	80	60.83	84.16	75	86.20
	Recall	9.00	10.40	8.70	8.00	7.30	10.10	8.00	10.20
Horse	Precision	100.00	97.50	62.50	85	91.67	94.16	90	95.20
	Recall	12.00	11.70	7.50	9.00	11.00	11.30	11.00	11.40
Mountain	Precision	39.17	29.17	40.00	65	44.17	50.00	60	60.05
	Recall	4.70	3.50	4.80	5.00	5.30	6.00	6.5	6.10
Food	Precision	52.50	76.67	64.17	55	45.00	92.91	75	94.9
	Recall	6.30	9.20	7.70	3.00	5.40	11.15	8.00	11.38
Average	Precision	68.25	72.16	61.58	76.5	65.75	78.54	78	83.5
	Recall	8.19	8.66	7.39	7.3	7.89	9.42	8.7	9.96