This paper proposes a novel framework for image splicing detection based on the inconsistency in the blur degree and depth information of an image. Firstly, the blur kernels of image blocks (local blur kernels) are estimated. Next, a multi-step reblurring technique is used to measure the relative blur degrees of the local blur kernels. The relative blur degrees are used as a feature to classify the image blocks based on the blur degrees. Furthermore, block-based and pixel-based techniques are incorporated for a fine segmentation of the regions with different blur degrees. Finally, the consistency in the blur degree and depth information of the regions are analyzed. Any inconsistency in the blur degree and depth information can be used as an evidence of image splicing. The experimental results show that the proposed approach can discriminate a wide range of blur degrees in natural and artificial blurred images which outperforms the state-of-the-art approaches. Furthermore, our proposed approach has been evaluated for image splicing detection. The result shows that it can be used successfully for detection of image splicing.
展开▼
