Improving Accuracy of Prediction with Forward Quantization Noise Method
In this work, simulation results are presented for detection of forgery in the JPEG compressed images.
Forward quantization method was simulated to determine the optimum value for the threshold for the
variance of the DCT coefficients. Two samples of Seam carving datasets were extracted for the
simulations. The threshold value for the variance of the DCT coefficients was varied from 0.005 to 0.0001
and it was found the accuracy of the detection of forged images is almost same for both the sampled
datasets. This method and accuracy of the forgery detection is independent on the dataset as the images
of the dataset are used only for testing purpose and not used in training or determining the values of
parameters of the model or approach.