项目名称: 基于小波域的纹理图像稀疏表示
项目编号: No.61272203
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 自动化技术、计算机技术
项目作者: 尤新革
作者单位: 华中科技大学
项目金额: 82万元
中文摘要: 本项目研究小波域纹理图像稀疏表示理论及相关方法,通过探讨小波构造与小波域上稀疏表示的关系,把分析具体纹理图像的矢量小波的构造条件作为约束项与图像稀疏表示模型中目标函数结合,研究适合具体纹理图像分析的矢量小波滤波器构造方法,解决适合纹理图像稀疏表示的自适应小波滤波器构造,建立基于矢量小波构造的小波域纹理图像稀疏表示模型,重点研究基于小波域稀疏表示模型的字典求解及快速算法,通过提出小波域稀疏表示方法解决不同尺度下纹理图像奇异结构的稀疏表示及不同结构纹理分离这些关键问题,突破传统小波变换和稀疏表示方法中对图像中不同结构纹理表示定位模糊、识别率低、模型求解复杂等局限,为不同结构纹理图像的表示与分析探讨新的理论工具和方法,形成新的基于小波域稀疏表示的纹理图像表示与分析理论体系。
中文关键词: 小波域;纹理图像;稀疏表示;;
英文摘要: In this project,We will study the theories and related methods of wavelet domain sparse representation of textural images. By exploring the relation between wavelet construction and wavelet domain sparse representation, the texture-specific vector wavelet construction conditions are combined with the objective functions of sparse representation models to construct vector wavelet filters which could be adapted to the specific textural image analysis tasks and to establish vector wavelet construction-based wavelet domain sparse representation models. The project focuses on wavelet domain-based sparse models and dictionary learning algorithms. Through investigating the key problems in wavelet domain textural image sparse representation models which obtain a sparse representation of textural images and separate different textural image structures at various scales, the limitations of traditional wavelets and sparse representation methods, e.g., vague position, low recognition rate, complex model and solution methods, are broken through. In all, this project aims at developing new theoretical tools and methods for representation of different textures and establishing a novel theoretical framework of wavelet domain sparse representation for textural image representation and analysis.
英文关键词: wavelet domain;texture image;sparse representation;;