Adaptive sparse coding on PCA dictionary for image denoising 基于PCA字典的图像去噪自适应稀疏编码.pdfVIP

Vis Comput DOI 10.1007/s00371-015-1087-x ORIGINAL ARTICLE Adaptive sparse coding on PCA dictionary for image denoising Qian Liu1 · Caiming Zhang1,2 · Qiang Guo1,2 · Hui Xu1 · Yuanfeng Zhou1 © Springer-Verlag Berlin Heidelberg 2015 Abstract Sparse coding is a popular technique in image Keywords Image denoising · Sparse coding · denoising. However, owing to the ill-posedness of denoising Iterative shrinkage · Principal component analysis problems, it is difficult to obtain an accurate estimation of the true code. To improve denoising performance, we collect the sparse coding errors of a dataset on a principal component 1 Introduction analysis dictionary, make an assumption on the probability of errors and derive an energy optimization model for image Image denoising is a fundamental and important issue in denoising, called adaptive sparse coding on a principal com- image processing because noise is inevitably introduced ponent analysis dictionary (ASC-PCA). The new method into images during acquisition and transmission. The noise considers two aspects. First, with a PCA dictionary-related may be characterized by several noise models, such as observation of the probability distributions of sparse coding Gaussian noise, salt and pepper noise and multiplicative errors on different dimensions, the regularization parame- noise. In this paper, we focus on additive white Gaussian ter balancing the fidelity term and the nonlocal constraint noise (AWGN) [3, 13]. Mathematically, the image denoising can be adaptively determined, which is critical for obtain- problem can be represented as solving the following equa- ing satisfying results. Furthermore, an intuitive interpretation tion: of the constructed