Existing　image　inpainting　networks　may　cause　distorted　image　structures　and　blurred　textures　during　restoration.　Therefore,　we　propose　an　image　inpainting　algorithm　based　on　Multi-Scale　Feature　Fusion　Generative　Adversarial　Network　(MSFF).　First,　we　use　a　Context　Fusion　Transformer　Block　(CFTB)　to　enlarge　the　receptive　field　of　the　defective　image　and　reduce　the　feature　loss　caused　by　sampling.　Second,　we　design　an　Aperture　Attention　Block　(ATAB)　to　capture　the　shallow　feature　information　and　perform　fine-grained　feature　fusion.　Finally,　through　the　decoder,　we　generate　high-quality　inpainted　images　step　by　step　from　coarse　to　fine.　The　results　of　multiple　experiments　show　that　our　method　has　a　better　inpainting　effect.　©　2024　John　Wiley　and　Sons　Inc.　All　rights　reserved.