Forensics of blurred images based on no-reference image quality assessment

The inexpensive hardware and sophisticated image editing software tools have been widely used, which makes it easy to create and manipulate digital images. The detection of forgery images has attracted academic researches in recent years. In this paper, we proposed a forensic method to detect globally or locally blurred images using no-reference image quality assessment. The features are extracted from mean subtracted contrast normalized (MSCN) coefficients and fed to SVM, which can distinguish the tampered regions from the original ones and can quantify the tampered regions. Experimental results show that this method can detect the edges of tampered regions efficiently.