基于JPEG块效应频域特性的合成图像检测

JPEG图像在压缩过程中所产生的块效应在功率谱曲线上体现为周期性波峰，而篡改JPEG图像所造成块效应不一致将导致周期性波峰的衰弱或消除。利用上述原理，提出了一种基于JPEG块效应频域特性的合成图像检测算法。算法对待测图像进行去噪，提取包含块效应的噪声，对其进行重叠分块并求得每块的块效应度量值，依据该度量值检测并定位篡改区域。实验结果表明，相对于传统的基于块效应不一致的算法，能够更好地检测多种不同图像格式的合成和篡改区域较小等情况。

Composite images detection based on frequency characteristics of JPEG blocking artifacts

The blocking artifacts of JPEG image produced during compression process exhibits as periodic peak of wave in power spectrum. The inconsistency of blocking artifacts induced by tampering JPEG image will weaken or eliminate the peak of wave. Utilizing the above principle, this paper proposes an algorithm to detect composite image based on the frequency characteristics of JPEG blocking artifacts. The noise of the detected image is removed to get noise. The noise is divided into overlapping blocks, and the blocking artifacts metrics of each block is calculated. The forgery regions are detected and located in terms of the metrics. Experimental results show that this algorithm performs better than the traditional algorithm based on the inconsistency of JPEG blocking artifacts in terms of various types of forgery with compositing differentiimage format and tamper with small area.