Image tampering detection using extreme learning machine

The use of textural descriptors were useful in the present work, which are calculated based on the Haralick features for image tempering using the Extreme Learning Machine (ELM). This method was applied to color images after converting it into YC_bC_r color system, and then the image is divided into blocks in order to apply the Local Binary Pattern (LBP) on each block belongs to each resulted color band. The textural features are then computed and encoded for the target image to be stored in a database file for that reconstructed image. The computed features enter the ELM classifier to carry out the processes of the training and classification. The training was performed on CASSIA-II dataset while testing was performed on CASSIA-I. The classification results gave a test accuracy of tempering detection about 99.7% when using the Y-band, 99.7% when using the C_b band, and 99.4% when using the C_r band. Whereas, the evaluation of the test results was good compared to previous work, this confirms the validity of the results and ensure the correct path of the proposed method.