Diffusion-based Semantic Image Editing with Edge-preserving Smoothing

Le résumé fourni par la source

Addressing the issues of uneven edges and semantic fragmentation in image editing using diffusion models, this paper proposes an improved method based on adaptive edge-preserving smoothing. This method significantly enhances the edge smoothness and semantic consistency of the edited regions by introducing an adaptive edge-preserving smoothing operation during the noise prediction stage of the diffusion model and combining it with a multi-level feature fusion strategy. Experimental results demonstrate that, compared with traditional methods, the improved method performs outstandingly in both qualitative and quantitative evaluations. Especially when dealing with tasks involving complex structures and large-scale deformations, it can generate more natural and high-quality edited results. Future work will focus on optimizing the algorithm efficiency and expanding its application in more image editing tasks.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.