EAGER: Image and Video Forensics: Detecting Image Manipulation by Content Analysis

复杂的照片编辑软件的出现使得处理数字图像变得越来越容易。目视检查往往不能明确地将伪造照片与真实照片区分开来。作为回应，法医技术已经出现，以检测由特定形式的照片处理引起的几何或统计不一致。PI渴望开发基于几何内容分析的新法医方法，其重点是发现照片中描绘的物体之间几何关系的不一致之处。所述二维图像中的几何关系对应于所述三维场景中存在的关系的投影；如果一个场景已知包含一个给定的关系，但投影关系在照片中不成立，那么人们可以得出结论，照片不是一个真正的场景投影图像。考虑到这一点，PI在这个探索性项目中的目标是构建一组必须在真实图像中满足的可测试约束，以便不满足的约束构成图像操作的确定的、客观的证据。这项工作的基本挑战包括：开发用于分析不完整照明信息的工具，建立皮肤反射率的可测试模型，考虑特征比较中的结构不确定性，以及建立法医图像分析的方法指南。更广泛的影响：该项目将创造客观检测图像操纵的工具，这将帮助记者、执法人员、科学家和其他人区分合法照片和伪造图像。本研究成果将通过学术出版物和在线源代码进行交流。与工业伙伴的合作也将使研究产生实际影响。

The advent of sophisticated photo editing software has made it increasingly easier to manipulate digital images. Often visual inspection cannot definitively distinguish the resulting forgeries from authentic photographs. In response, forensic techniques have emerged to detect geometric or statistical inconsistencies that result from specific forms of photo manipulation. The PI aspires to develop new forensic methods based on geometric content analysis, which focus on finding inconsistencies in the geometric relationships among objects depicted in a photograph. The geometric relationships in a 2D image correspond to the projection of the relations that exist in the 3D scene; if a scene is known to contain a given relationship but the projected relation does not hold in the photograph, then one may conclude that the photograph is not a true projective image of the scene. With this in mind, the PI's goal in this exploratory project is to build a set of testable constraints that must be satisfied in real images, so that an unsatisfied constraint constitutes definitive, objective evidence of image manipulation. Fundamental challenges of this work include: developing tools for analysis from incomplete lighting information, building testable models of skin reflectance, accounting for structured uncertainty in feature comparison, and establishing method guidelines for forensic image analysis.Broader Impacts: This project will create tools for objectively detecting image manipulation, which will help reporters, law enforcement, scientists, and others differentiate between legitimate photographs and forged images. The products of this research will be communicated via academic publications and online source code. Collaborations with industrial partners will allow the research to have practical impact as well.