The continual advancement of image editing techniques has made manipulated images easier to create. Improper use may lead to the proliferation of forged images. In order to detect and locate forged regions within forged images, existing research utilizes various feature views to capture subtle forgery traces. However, forged images exhibit complex higher-order relationships, such as group interaction among regions. The interaction reflects inconsistencies among regions. Therefore, we propose a novel Dual Hypergraph Convolution Network (DHC-Net) to enhance the localization of forged regions by representing group interactions using hypergraphs. The DHC-Net constructs region-wise and edge-wise hypergraph convolution branches to refine the localization of forged region. We validate the DHC-Net on four widely used public datasets, including CASIA1.0, NIST, Columbia, and Coverage. The results demonstrate that the proposed DHC-Net achieves advance localization accuracy.