LAUSR

With the ever increasing use of digital media, image tampering has become imperative. This spurs the need to identify such tampering for authentication and jurisdiction. The main idea of this paper is an assessment of the possible light source direction from the image. This technique uses the inconsistencies in the light source direction to detect the image forgery. Initially, in the preprocessing step on input image, surface normals are calculated using surface texture profile. RED band is mainly used for obtaining surface texture information and, further, surface normal calculations are done. With estimated illumination profile and normals, the incident angle θi is computed for various chosen image patches. The θi angle is the estimated angle from image object to light source direction. The inconsistency in θi values is used as an evidence of tampering. The proposed technique is tested on different known fake images and is found capable of identifying manipulated objects in an image. This technique works for homogenous illuminated surfaces and has better forgery detection accuracy. Additionally, our technique also diminishes human intervention for forgery detection. The performance of proposed forgery detection technique is examined using CASIA1 image database to give users a feel of the performance.