Modeling Small-Scale Patterns in Natural Images by Sequential Eigenvalue Problems in Sobolev Spaces
-
摘要: 提出了一种描述自然图像小尺度模式的新方法. 其中心思想是将小尺度信息的建模问题转化为图像区域上Sobolev空间的序列能量极小化问题, 进而转化为序列特征值问题. 从数学上分析了小尺度模式的多层次结构及模型的收敛性. 本文还提出一种新的自适应多层次化图像表示方法, 并可应用于图像合成及视觉感知等方面. 从数值计算的角度上, 通过稀疏对称矩阵特征值分解可方便地获得不同层次的小尺度模式.Abstract: This paper develops a new framework to characterize small-scale patterns. The core idea is to formulate the problem of modeling small-scale patterns by sequential minimizing energy functionals in Sobolev spaces, leading to a sequence of eigenvalue problems. Mathematical analysis on the structure of the small-scale pattern at each level and the convergence behavior of the proposed model are explored in detail. We obtain a novel effective, adaptive, and hierarchical image representation. And the model will benefit applications such as image synthesis and computational visual perception. Numerically, small-scale patterns can be easily computed by eigenvalue decompositions of sparse, symmetric matrices.
-
Key words:
- Image modeling /
- small-scale patterns /
- visual perception /
- calculus of variations /
- Sobolev space /
- eigenvalue problem
点击查看大图
计量
- 文章访问数: 1664
- HTML全文浏览量: 63
- PDF下载量: 941
- 被引次数: 0