Saliency detection as a support for image forensics

In this work we introduce a new application of saliency detection as a support for forensic analysis in digital images. Indeed, although there are some attempts to achieve both detection and localization of image tampering, the majority of forensics tools require hypothesis about the modified area. Since we claim that salient objects/subjects, conveying the semantic content of the image, are the regions whose integrity is more critical, we propose to combine salient object detection and forensic analysis on the corresponding output. To this aim we introduce here an improved version of a saliency map extractor based on segmentation and describe the application of a tampering detection method allowing digital composite detection. Experimental results are presented and discussed. © 2012 IEEE.