Comparison of Scale Invariant Feature Transform and Speed Up Robust Feature for Image Forgery Detection Copy Move

摘要

Copy move is one of the techniques used to fake images. Many studies have developed methods to detect displacement of counterfeit copies of images. Some researchers use methods for the process of extracting features in an image such as SIFT and Speed Up Robust Feature. This study aims to compare the results of the detection of copy move image forgery using the SIFT algorithm and Speed Up Robust Feature. Testing will be done is in terms of accuracy and execution time. At the pre-processing stage, the image decomposition process is done using the Discrete Wavelet Transform (DWT) method. In this research there are 5 types of testing, namely without transformation, scale transformation, rotation transformation, lighting transformation, and blur transformation. Modeling the program code will use the MATLAB programming language. By using the transformation process, there are more key points, but fewer key points. Based on testing that has been done that the SIFT algorithm has a higher accuracy than SURF. The difference in the 1st test was 16.32%, the difference in the 2nd test was 9.96%, the difference in the 3rd test was 13.32%, the difference in the 4th test was 13.44%, and the difference in the 2nd test 5 of 1.87 %. To test the execution time, the SURF algorithm has a faster time compared to the SIFT. The difference in testing 1, 2, and 3 amounted to 1.97 seconds, the difference in the 4th test was 0.7 seconds and the difference in the fifth test was 0.62 seconds.