Noise Resilient Local Gradient Orientation for Content-Based Image Retrieval

Table 6

Performance of NRLGO at Corel dataset with 1000, 5000, and 1000 images at different combinations of color, gradient orientation, and local contrast information vector.

Dataset	Performance	Color	Texture	Contrast	Texture + contrast	Color + texture	Color + contrast	Proposed
Corel 1k	Precision (%)	76.8	54.6	52.8	59.4	79.5	78.25	83.5
	Recall (%)	9.02	6.38	6.10	6.90	9.22	9.30	9.96
Corel 5k	Precision (%)	60.2	27.5	31.5	39.5	61.6	61.8	65.8
	Recall (%)	7.15	3.06	3.89	4.50	7.30	7.25	7.89
Corel 10k	Precision (%)	48.5	20.5	25.9	28.6	50.7	49.3	53.07
	Recall (%)	5.51	2.30	3.05	3.45	5.80	5.95	6.36