Indian Journal of Science and Technology

Abstract

Objectives: The image forgery detection is the technique in which pixels are marked in the image, which are not similar to other pixels of the images. The Principal Component Analysis (PCA) is the classification of neural networks which will analyze each pixel of the image and classify pixels according to pixel type. Method: The PCA algorithm takes training and trained dataset as input and drive new values according to input image. In the paper improvement in the PCA algorithm with usage of Scale-Invariant-Feature-Transform algorithm (SIFT Algorithm), is proposed for image-forgery. The SIFT algorithm is the algorithm which analyze each pixel of the image and define type of pixels in the image. The output of the SIFT algorithm is given as input to PCA algorithm for data classification. The PCA algorithm will classify the data according to SIFT algorithm output. Findings: The results demonstrate that the proposed algorithm executes well in terms of “Peak Signal-to-Noise Ratio” (PSNR), “Mean-Square-Error” (MSE), fault detection rate and accuracy value.
Keywords: Forgery Detection, MSE, PCA, PSNR, SIFT