Robust Spatial Domain Features for Detecting Spliced Image

Digital image forgery (DIF) detection has gotten a lot of attention lately. One common method used to generate image forgery is image splicing. Image splicing produces manipulated images that are difficult to recognize quickly. There is potential for improvement because the most of currently used spliced image detection methods have poor detection accuracy. Thus, this paper presents a spliced image detection (SID) technique relies on Gabor filter. Three steps are involved in the proposed image splicing detection method: (1) One type of image pre-processing is the transformation of RGB images. (2) Using the Gabor filter, we will now extract the final distinguishing feature. Gabor's primary benefit is in its capacity to obtain texture information from the source image. (3) Lastly, an Adaboost, KNN, and Naive classifier are trained for classify the image into authentic or spliced using the resultant feature. When tested on the CASIA v2.0 spliced image forgery database, the proposed approach of integrating a Gabor features with Adaboost classifier has an accuracy of 98%. A comparative study reveals that the proposed Gabor-Adaboost outperforms the most current splicing detection methods.