图像锐化检测关键技术研究

Currently, a vast number of fake images have appeared in the Internet, which has done great harm to the positive dissemination of information. Thus, how to verify image authenticity and detect image tampering is a technical problem that needs to be solved urgently. Image sharpening detection, being a technique of image forensics, is one of the important ways to find the trace of image tampering. In recent years, although a certain progress has been made in the field of image sharpening detection, there are still some unsolved issues, including the theoretical analysis of image sharpening operation in the 2-dimensional (2-D) case, the 2-D characterization of overshoot effect, and the parameter estimation of image sharpening. This project mainly studies the fundamental theories and methods of image sharpening detection. In order to address the above-mentioned issues, we develop a mathematical model to describe the process of unsharp masking sharpening from a 2-D point of view, then propose a sharpening detection method based on 2-D overshoot artifacts and a sharpening parameter estimation method based on original value regression. Finally, with the help of the above studies, we present an image tampering detection technique by combining the non-overlapping partitioning and the sliding partitioning, which might provide reliable technical support for the practical application of image forensics.

当前，网络上充斥着大量的虚假图像，给信息的健康传播造成了严重危害。因此，如何鉴别图像真假以及检测图像篡改是当前亟需解决的一个技术问题。图像锐化检测，作为一种数字图像取证技术，是发现图像篡改痕迹的重要技术手段之一。近年来，图像锐化检测的研究已取得了一定进展，但依然有些重要问题还未得到很好地解决，包括：缺乏图像锐化机理在二维情况下的理论分析；缺乏有效的过冲效应的二维刻画方法；尚未开展锐化参数估计方面的工作。本项目主要研究图像锐化检测的基础理论与方法，针对上述问题，首次从二维角度对掩模锐化过程进行数学建模，提出基于二维过冲痕迹的锐化检测方法和基于原值回归的锐化参数估计方法；在此基础上，提出非重叠分块与滑动分块相结合的图像篡改检测技术，为图像取证在实际中的应用提供可靠的技术支持。

当前，网络上充斥着大量的虚假图像，给信息的健康传播造成了严重危害。因此，如何鉴别图像真假以及检测图像篡改是当前亟需解决的一个技术问题。为解决这一问题，本项目围绕图像锐化检测等数字图像取证技术开展了一系列研究工作，主要包括图像锐化检测、图像重采样率估计、JPEG压缩量化步长估计、图像copy-move伪造检测、图像copy-move来源/目标辨别、JPEG域图像增强取证分析、图像扩散修复检测、图像拼接定位、以及deepfake取证/反取证等。在上述研究的基础上，发表（含在线发表）国际期刊和会议论文共26篇，其中国际SCI权威期刊论文24篇、IEEE/ACM Transactions论文17篇；申请发明专利12项。