项目名称: 基于矩阵分解的图像表示方法及其应用研究
项目编号: No.61502506
项目类型: 青年科学基金项目
立项/批准年度: 2016
项目学科: 其他
项目作者: 肖延辉
作者单位: 中国人民公安大学
项目金额: 20万元
中文摘要: 图像表示是图像处理和模式识别领域中的核心研究问题之一。一个有效的图像表示方法不仅有助于挖掘图像潜在的数据结构,而且有利于降低数据存储和传输的成本。非负性、稀疏性、鲁棒性以及判别性是图像表示理论的核心问题,本课题拟从矩阵分解的角度出发,围绕图像表示方法中的矩阵非负分解和稀疏分解两个重要的研究方向展开研究,主要包括:判别的图像低维表示方法,基于全局最优解的非负矩阵分解方法,鲁棒的图像稀疏表示方法等。本项目的研究成果可以广泛应用于图像分类和聚类、图像检索、人脸识别、目标检测、图像去噪等领域,具有重要的理论意义和实用价值。
中文关键词: 图像表示;非负矩阵分解;稀疏表示;图像分类
英文摘要: Image representation is a fundamental problem in image processing and pattern recognition tasks. A good representation can typically reveal the latent structure of data, and further facilitate these tasks in terms of learnability and computational complexity. This project will make a study of the image representation of such non-negativity, sparseness, robustness and discrimination. For such purpose, the matrix decomposition theories including non-negative matrix factorization, sparse coding, dictionary learning, sparse auto-encoders and independent component analysis are utilized. Specifically, we will primarily study the following subjects: the discriminative non-negative matrix factorization for image representation, the globally optimal solution to non-negative matrix factorization, and the robust sparse representation for image classification and face recognition. The research achievements of this project can be extensively applied to the fields including image classification, image clustering, image retrieval, face recognition, object detection and image denoising, and have great significance in both theory and practice.
英文关键词: image representation;non-negative matrix factorization;sparse representation;image classification