Interactive Segmentation of Art Images Based on Hypergraph Convolutional Networks

Existing image segmentation algorithms often suffer from a lack of interactivity and a low level of segmentation accuracy. To solve this issue, this paper proposes the novel interactive segmentation of art images based on the Hypergraph Convolutional Networks (HCN). In order to further improve the accuracy of the image segmentation, this paper applies the interactive method based on center point correction to image segmentation based on graph convolutional neural network, and proposes a new interactive image segmentation method: based on graph convolutional neural network an interactive segmentation method for networks. In the designed model, 2 major ideas are provided, namely: (1) According to the network structure of U-Net, this paper builds an encoder by stacking coding unit blocks, and then the coding part of HGU-Net is composed of multi-layer encoders. (2) An interactive modification method is proposed to correct segmentation results in real time by adding new indicator strokes. Through the comparison testing with CNNs and U-Set, the performance is validated.