MII-DFNet: A Medical Image Inpainting Framework Based on Deep Fusion Network
Yutong Yuan, Sizhe Dai · 2023
Since medical images are easily captured with missing areas due to uncertainties such as instrument or human factors, which can seriously affect doctors' diagnosis and patients' treatment, medical image inpainting techniques have become very important. How to make the restored images not only rich in structural and textural information, but also smooth transitions in the boundary regions to blend into the adjacent regions of the missing parts has always been a research focus in the field of computer vision. In order to address the above problems, this paper proposes a medical image inpainting framework (MII-DFNet) based on deep fusion network, through which the medical images covered by mask are restored to complement the original information. And two common medical images, CT and MRI, are used as our test datasets to verify the performance of the model. The experiments show that our method is better in restoration and generates higher quality images.