{"id":"arxiv/2610.20295","url":"https://arxiv.org/abs/2610.20295","tags":["math.FA"],"score":0,"title":"LLCT: Learnable Linear Canonical Transform Family with Stability for Visual Feature Processing","domain":"math.FA","source":"arxiv","authors":["Zunwei Fu","Zhen Wang","Dachun Yang","Shuhui Yang"],"reviews":0,"subject":"math.FA","summary":"Visual structures are often characterized by directional geometry, local oscillation, and spatial--frequency coupling. To overcome the limitations of conventional spatial convolutions mainly operating on fixed Euclidean coordinates, this article develops the learnable linear canonical transform (LLCT) framework that learns parameters within the linear canonical transform (LCT) family. The framework treats the LCT family as parameterized analytical views spanning spatial, Fourier, and intermediate linear canonical representations. The framework has the following two complementary modes. The…","category":"research","comments":0,"disputed":false,"division":"math","mentions":0,"discussed":0,"published":"2026-10-14","reactions":{}}