基于混合特征提取的图像来源鉴别算法  

Image Source Identification Algorithm Based on Mixed Feature Extraction

作　　者：黄曜[1] 许华虎 欧阳杰臣 高珏[1] 

机构地区：[1]上海大学,上海200444 [2]上海上大海润信息系统有限公司,上海200444

出　　处：《计算机技术与发展》2016年第4期11-15,共5页Computer Technology and Development

基　　金：国家重大科技专项课题(2009ZX04001-111)

摘　　要：随着数码图像的普及,图像盲取证成为时下的研究热点之一。如何识别图像来源是其中主要的研究内容。特征提取是对图像进行鉴别的前提。文中通过对现有的特征提取方法进行研究,针对现有单一特征提取不能完全反映图像特质导致识别出错的问题,提出混合特征提取的概念,依次提取图像的颜色特征、纹理特征以及统计特征,从而提高图像来源的识别率。通过实验对文中提出的算法进行了验证。结果表明,文中提出的混合特征提取算法较任一单一特征提取算法都能取得更好的图像来源鉴别率。With the popularity of digital images,blind image forensics has become one of the hotspots nowadays. The main research content of blind image forensics is how to identify the image source. Feature extraction is a prerequisite to identify the image. By studying the existing feature extraction methods,aiming at the problem that the single feature extraction may not fully reflect the image characteristics to lead to the recognition error,the concept of mixed feature extraction is proposed,extraction of the features of color,texture and statistics to improve the recognition rate for image source. The algorithm proposed in this paper is validated through the experiment. The results show that the mixed feature extraction algorithm proposed can achieve better image source identification rate compared with any single feature extraction algorithm.

关 键 词：图像盲取证 单一特征提取 混合特征提取 图像来源鉴别率 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]