EURASIP Journal on Advances in Signal Processing

Table 4 Average classification accuracy, i.e., 1 - (FAR + FRR)/2, in percent with various values of QF₁and QF₂for patch size of 64 × 64

From: Detection of shifted double JPEG compression by an adaptive DCT coefficient model

QF₁	QF₂
	50	60	70	80	90
50	51.86/50.40/50.96	51.55/55.15/53.51	54.89/70.08/62.07	56.32/78.95/70.02	62.29/82.12/76.71
50	50.90/58.28	51.56/71.32	54.29/86.77	56.58/91.65	63.01/96.66
60	51.31/50.26/49.34	50.57/50.87/52.10	51.40/55.24/56.21	53.92/70.50/62.51	59.02/78.39/71.51
60	51.95/54.05	50.34/56.96	51.75/69.80	54.31/90.71	60.89/92.17
70	50.62/50.44/50.70	50.33/50.55/51.09	51.30/51.13/51.07	52.71/59.49/60.02	56.33/70.25/65.25
70	50.80/51.10	49.60/52.04	50.45/59.19	52.63/71.76	57.22/92.12
80	50.42/50.92/50.47	49.62/50.63/50.98	51.62/50.42/50.09	50.49/51.91/51.77	54.34/58.41/59.39
80	49.79/50.83	49.77/50.44	49.66/53.49	50.33/56.28	53.90/75.50
90	50.44/50.27/49.26	49.86/49.41/49.20	49.89/49.91/49.45	49.39/50.37/49.62	50.02/50.14/50.48
	49.13/51.70	50.64/50.39	49.87/50.45	50.48/51.68	50.26/58.96

Average classification accuracy, i.e., 1 - (FAR + FRR)/2, in percent, with various values of QF₁ and QF₂ for patch size 64 × 64. The accuracies a1/a2/a3/a4/a5 in the table are obtained by Luo's (a1), Bianchi I's (a2), Bianchi II's (a3), Chen's (a4), and the proposed method (a5), respectively. The best accuracies among the five methods investigated are boldfaced. Note that an accuracy of 50 is equivalent to random guess.

Back to article page