Abstract: In the image splicing process, feathering edge of the splicing area are different from the natural image edge. And these differences can be used to detect the image splicing region. However, when some edges in the image are similar to the features of feathered edges, the method of directly extracting the features of feathered edges can lead to test result incorrect. If it is possible to expend the difference between the edges of the spliced area and the edges of the natural image, detection of tampered images will become more accurately. Therefore, this paper proposes a new edge detection method about the image of feathered edges based on USM enhancement. Firstly, the difference between feathering edge and natural edge feature in the splicing image is enlarged by using the method of UMS enhancement. Then, the edge pixel feather radius of image is calculated. Finally, the edge pixels with similar radius are found, and the splicing region is located in the image. The experimental results show that the splicing region can be detected more accurately when there are edges similar to the feathering edge in the spliced image.

Key words: edge feather, USM enhancement, feather radius, splicing image detection

摘要： 图像拼接过程中，经过羽化模糊处理的图像边缘特征是不同于自然图像边缘特征的。因此，可利用图像边缘羽化特征来进行图像拼接区域的检测。然而，当图像中某些边缘与羽化边缘特征相似时，直接提取边缘羽化特征的方法会导致误判。如果能够扩大拼接区域的边缘与自然图像边缘之间的差异，则可以更准确地检测出图像中的篡改区域。因此，提出了一种基于USM增强的边缘羽化拼接图像检测方法。首先，对待检测的拼接图像进行UMS增强，扩大拼接图像中羽化边缘与自然边缘特征的差异性。然后，计算增强后图像边缘像素的羽化半径。最后，寻找半径相似的边缘像素，定位图像中的拼接篡改区域。实验结果表明，在拼接图像中存在与羽化边缘相似的边缘时，该方法能更准确地检测出拼接区域。

关键词: 边缘羽化, UMS增强, 羽化半径, 拼接图像检测