Generalized Non-locally Centralized Image De-noising Using Sparse Dictionary

What is it about?

Recently, image de-noising algorithm based on sparse representation has received an increasing amount of attention. Such algorithms proposed a comprehensive sparse representation (CSR) model, by solving the sparse coding problem and choosing the proper method for dictionary updating to achieve better de-noising results. Therefore, the construction of learning dictionary has become one of the key problems that limit the de-noising effectiveness. The non-locally centralized sparse representation de-noising algorithm uses principal component analysis (PCA) method to achieve dictionary updating. Nevertheless, the instability of a single complete dictionary in sparse coding leads to erratic result in the process of the original image restoration. In this paper, we present a new method named generalized non-locally centralized sparse representation algorithm (GNCSR). In the proposed method, we cluster the training patches extracted from a set of example images into subspaces, and then train dictionaries for subspaces by sparse analysis K-SVD (k-singular value decomposition) dictionary, which is utilized to construct coded sub-block dictionary to avoid the instable results caused by a single dictionary. Experiments show that the improved method has better signal noise ratio and de-noising effect compared with other methods.

Featured Image

Why is it important?

The main objective of this work is to reconstruct a dictionary algorithm for image denoising. Due to the degradation of the observed noisy image, the sparse representations by conventional models and dictionary learning may not be accurate enough for a faithful reconstruction of the original image. In this paper, we first present an improved de-noising method that is NCSR (Nonlocally centralized sparse representation) algorithm. Based on the NCSR, we cluster the training patches extracted from a set of example images into subspaces, and then training dictionaries for subspaces by sparse Analysis K-SVD (K-Singular Value Decomposition) method, which adopts adaptive multiples the signal for each cluster and achieves better parameter estimate. During the dictionary update step, alternating sparse-coding and dictionary update steps for a fixed number of iterations, thus leading to a more stable and sparser representation and better image de-noising results. Experiments on synthetic and natural images show that our method behaves well in image de-noising performance.

Perspectives