基于无限光源模式的数字图像盲鉴别技术

Blind identification technology for digital images based on infinite light source model

图像篡改通常会破坏自然图像的光照一致性,根据这一特点将光源方向的不一致性作为检测图像篡改的依据。针对在无限光源照射下生成的图像,根据图像的测量强度与计算强度间的误差函数和无限光源模式的约束条件函数,用Hestenes-Powell乘子法计算图像中不同对象和背景的光源方向,并以此判断图像是否被篡改。实验结果表明:本文算法能够有效地计算出无限光源模式下图像中不同对象和背景的光源方向,其正检率已经达到了82.3%。

Inconsistency in light source direction is taken as the basis for the detection of image tampering. Focusing on the images generated under the irradiation of infinite light source, the Hestenes-Powell multiplier method is used to calculate the light source direction of different objects and backgrounds in an image and to determine whether the image has been tampered with. The calculation is in accordance with the error function between the measured intensity and calculated intensity, and the constraint function suitable for infinite light source model. Experiment results show that this algorithm can effectively calculate the light source direction of different objects and backgrounds in the image lased on infinite light source model, and the detection rate reaches up to 82.3%.