Less is More: Efficient Image Vectorization with Adaptive Parameterization

b) O&R [12] (a) Raster Image (e) Image Editing (d) Ours (c) SGLIVE [40] Figure 1. Given a raster image as input (a), typical image vectorization methods, i.e., O&R [12] (b) and SGLIVE [40] (c), mainly rely on preset parameters (i.e., a fixed number of paths and control points), and fail to produce pleasant results when the image structure is complex. In contrast, our method (d) can perform well with an adaptive number of paths and control point parameters based on the input image complexity, resulting in fast computation speed, high vectorization accuracy, and flexible editing applications (e), e.g., image color adjustment, object animation, and icon customization.