Edge-based Blur Metric for Tamper Detection

Abstract

In order to check the aboriginality and integrity of a digital photograph, a blind forensics scheme for detecting blur manipulation is proposed in this paper. A cost- effective local blur estimator is designed to measure the blurriness of each pixel along a doubted edge. Consistency metric of such a blurriness sequence is constructed based on the deviation from its linear fitting. Then the metric is used as evidence for identifying blur operation. Experimental results both on synthetic and natural images have shown the efficiency of our proposed blur forensics scheme.