基于Benford模型的自然图像与计算机生成图像的鉴别

针对目前自然图像和计算机生成图像的鉴别方法鉴定准确率不高的问题,提出了一种基于Benford模型的自然图像与计算机生成图像的鉴别方法.本算法利用DCT域AC系数首位有效数字的Benford曲线分布,分别对图像的RGB三个色彩通道进行统计,以3条概率分布曲线的重合程度作为鉴别取证的依据,对2类图像进行正确分类.实验结果表明,该方法可有效地鉴别自然图像和计算机生成图像,与已有算法相比具有更高的识别率,鉴别准确率达97.17%,且计算量小、易于实现,为图像取证、数字防伪鉴别等提供可靠的依据.

Distinguishing Computer Graphics From Natural Image Based on Benford Model

An improved method with a higher identifying accuracy rate is presented to identify natural images and computer graphics,and the new detection scheme of computer graphics and natural images is based on the Benford model.This algorithm separately calculates the three RGB color channels by probability distribution of the first significant digit of AC coefficient in discrete cosine transform(DCT) domain and correctly classifies the two types of images based on fitting degree of the three curves.Experimental results show that this method can effectively identify natural images and computer graphics.Compared with the existing algorithms,this algorithm has a higher recognition rate,which makes calculation less and easier to implement,and the identifying accuracy rate comes to 97.17%.This provides credible supports for some relative domains such as image forensics and digital anti-fake identification.