Abstract: This paper proposes a new model for image decomposition to achieve staircase reduction. The energy for the cartoon interpolates between total variation regularization and isotropic smoothing, while the energy for the texture interpolates between Meyer’s G norm and H-1 norm. These energies are dual in the sense of the Legendre-Fenchel transform and their adaptive behavior preserves key features such as object boundaries and textures while avoiding staircasing in smooth regions.

Key words: image decomposition, total variation minimization, dual norm, texture

摘要： 提出一种减少阶梯现象的图像分解模型。该模型所表示的结构成分的能量介于全变差正则化和各向同性光滑化之间，纹理成分所表示的能量介于Meyer的G范数和H-1范数之间，它们在Legendre-Fenchel变换的意义下是对偶的，根据图像的局部信息自适应地调整。实验表明，新模型能很好地避免在光滑区域出现的阶梯现象，有效保护图像的边缘和纹理信息。

关键词: 图像分解, 全变差最小化, 对偶范数, 纹理

CLC Number: 

• TP391