基于图像稀疏特性的图像表示、编码与重建研究

项目名称： 基于图像稀疏特性的图像表示、编码与重建研究

项目编号： No.61472011

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 计算机科学学科

项目作者： 刘家瑛

作者单位： 北京大学

项目金额： 80万元

中文摘要： 随着多媒体技术的迅猛发展和互联网社交网络的广泛兴起，数字图像作为多媒体信息中最重要的视觉信息载体，逐渐给人们生产生活和信息沟通方式带来巨大的变革。当前数字图像采集设备的普及对数字图像处理提出了更高的挑战。如何根据图像固有性质和人类视觉特性寻求高效的图像表示、建模及重建方法，具有十分重要的意义。本项目将围绕图像稀疏特性对图像表示建模、稀疏编码和重建应用展开深入的研究，通过分析新兴媒体领域图像高层语义和底层结构信息，重点研究和实现结构化图像稀疏表示模型；并根据图像上下文信息，提出高效鲁棒的稀疏编码方法，以提高图像稀疏重建的有效性和鲁棒性；最后根据对稀疏表示模型在不同图像重建领域的应用特性分析，设计和实现一种基于图像稀疏特性的统一重建框架。本项目的研究，对于稀疏表示理论的研究，建立基础研究与应用平台，增强图像稀疏表示模型在不同领域的扩展性，进一步推动图像稀疏表示理论和方法的发展具有重要的意义。

中文关键词： 稀疏表示；图像处理；图像去噪；图像去模糊；图像增强

英文摘要： With the rapid development of multimedia technology and fast evolvement of internet social network, digital image, as the most important visual information carrier among the multimedia information, has brought huge revolution to the human's life and communication styles. In recent years, the popularity of digital image capture devices puts forward higher challenge to digital image processing. Therefore, finding more efficient theories and technologies for image representation, modeling and image reconstruction has become a very important task in the image processing research field. In this project, we will focus on the research on image representation, sparse coding and reconstruction applications based on the image sparsity property. In sparsity-based image representation and modeling, we mainly incorporate the high level semantic information and low level structural information to the basic model for structural image sparse representation; In image sparse coding and decomposition, we mainly focus on the context-aware image information to improve the effectiveness and robustness of sparse coding methods. In sparsity-based image reconstruction, we will study the similarities and differences among different applications, then design and implement an unified framework of image reconstruction based on image sparsity property. This project is of great importance to the research on sparse representation and techniques, building basic research and application platform, extending the scalability of sparsity model and promoting the development of sparse representation theories and techniques.

英文关键词： sparse representation;image processing;image denoising;image deblurring;image enhancement