A New Approach for Large-Scale Scene Image Retrieval Based on Improved Parallel -Means Algorithm in MapReduce Environment

Table 2

Contrast of the retrieval accuracy (%) of different methods based on COREL datasets.

Dataset	Retrieval precision rate (%)
	-Means	-Means	Global -Means	-Means++	Parallel -Means	The proposed method
Data 1	90.79	91.42	91.57	92.31	94.46	98.71
Data 2	88.59	89.83	89.81	90.62	93.83	98.03
Data 3	87.87	88.14	89.01	89.54	93.76	98.00
Data 4	85.83	85.96	86.25	87.49	92.17	96.57
Data 5	84.81	85.07	85.49	85.91	92.17	96.56
Data 6	82.35	83.89	83.96	83.97	90.73	94.99
Data 7	79.64	81.44	81.97	82.35	90.06	94.55
Data 8	74.91	76.87	77.28	77.40	89.45	93.76
Max	90.79	91.42	91.57	92.31	94.46	98.71
Min	74.91	76.87	77.28	77.40	89.45	93.76
Mean	84.35	85.33	85.67	86.20	92.08	96.33
Standard deviation	4.87	4.39	4.34	4.57	1.75	1.70