Spatial-contextual texture and edge analysis approach for unsupervised change detection of faces in counterfeit images

Abstract

Images transfer more information about expressing an occasion than words. The skillful knowledge should exist to perceive the change in objects present in those images. Also, there is no sign existing to crisscross the objects are liable or not. The internet images should be legitimate to approve its truthiness. The cloned image forgery can be performed by many sophisticated cameras and by editing software. Therefore, a proposal should present to detect the change in counterfeit or forgery-based images to warrant the objects existing in the image are true. The input images are pre-processed to generate the Gray world and illuminant maps. The faces in the input images are segmented using both automatic and semi-automatic methods. The texture from the faces is bred using the values from pixels of locality. Canny detector is applied to distinguish the edge in the face. The features acquired from the faces are compared with each other to sense the counterfeit face that is spliced in the original image to make them as the composite image. The features are then trained to catalog them as forgery or no forgery. Existing methodologies has the capability to identify forgery in image with the extreme of two faces. The proposed method has the prospect to spot every faces present in the image. The cloned counterfeit faces are removed from the image by spatial contextual correlation strategy of image completion. Experimental results show that the proposed methodology achieves well than the other approaches present in the literature.