关键词: 拼接伪造图像, 光源颜色, 噪声

Abstract: Image splicing is a common image content tampering operation. It is still a difficult task to detect splicing areas in image forensics. Most of the existing splicing detection methods detect tampered areas based on the inconsistency of an image characteristic, while the actual splicing forgery often causes the change of a variety of image characteristics. Aiming at the problem that the single feature extraction can not fully reflect the image characteristics to lead to the low detection accuracy, this paper proposes a new method for detecting the image splicing efficiently by extracting the mixed features of the illuminant color and noise in the local regions. The experimental results show that the mixed feature extraction method can achieve higher accuracy than single feature extraction method. 

Key words: splicing image, illuminant color, noise