基于特征和稀疏性的图像去噪理论和算法研究

The problem of image denoising is a common and a key technology in the field of video processing, computer graphics, computer vision, clinical medicine and space science. It’s the research hotspot in these areas, and is a fundamental and difficult problem. So it has the extensive application background. This project demonstrate research on the theories and algorithms involved in the process of generating a clear image from noisy image. The main research contents include: basic theories of image denoising, noisy image segmentation and sparse representation and high quality image generation. This project uses the features and sparse representation of the image as the breakthrough point. Additionally it uses the similarity measure of the image block and edge extraction as well as the fast calculation of the singular value as the main line and the key of research, to carry out a series of innovative research on the image denoising. The research brings out three new models, which are the similarity measure of the image block, the edge extraction based on the features, and the fast calculation of singular value. The research involves the theory, algorithm and technology of image denoising and the open prototype system. The main goal of the project is to provide new theories and methods for image denoising. Furthermore it provides new ideas and a series of efficient and robust methods in the field of image segmentation, edge extraction image denoising and image enhancement. Therefore, to achieve the expectation goal of the project is of great theoretical significance and important value of application.

图像去噪是视频处理、计算机图形学、计算机视觉、临床医学和空间科学等领域的共性关键技术，是这些领域的研究热点，也是急需解决的基础问题和难点问题，有广泛的应用背景。本项目研究由含噪声图像生成清晰图像过程中所涉及的理论和算法，主要研究内容包括：图像去噪的有关基础理论；含噪声图像分割和稀疏表示；高质量图像的生成。项目以图像的特征和稀疏性表示为切入点，以图像块的相似性度量、边缘提取和奇异值的快速计算为研究的主线和关键，开展图像去噪的系列创新研究，提出图像块的相似性度量、基于特征的边缘提取和奇异值的快速计算三个新模型。研究内容涉及图像去噪方面的理论、算法与技术以及原型开放系统。项目的主要目标是为图像去噪问题的解决提供新理论和方法，为上述领域中的图像分割、边缘提取、图像去噪和图像增强等问题的解决提供新思想和一系列高效鲁棒的方法。因此，项目预期目标的完成具有很好的理论意义和重要的应用价值。