Image Splicing Localization Using Superpixel and Wavelet Mean Squared Error

Image splicing is a form of image forgery where one image is seamlessly pasted onto another. While image splicing itself is not necessarily illegal and is often used for aesthetic purposes, entertainment, or humor, it can also be misused for criminal activities. To aid digital forensics in detecting the specific location of the spliced image area, an image splicing localization algorithm is required. Previous research has proposed a method for detecting the splicing area by employing superpixel segmentation and noise level estimation. This method involves dividing the photo into smaller parts, estimating the noise level in each part, and clustering these estimated noise levels. By clustering the noise levels, the different clusters represent distinct levels of noise and help identify the splicing area. In an effort to enhance the previous research, this study introduces a new method for detecting the splicing area using superpixel segmentation and wavelet mean squared error. The research demonstrates that the wavelet denoising mean squared error value can be effectively employed to detect the image splicing area, outperforming the noise level estimation method under certain conditions.