Dictionary Learning for SAR Images Despeckling: A Comparative Study

In recent years, dictionaries combined with sparse learning techniques became extremely popular in computer vision. The image denoising approaches can be categorized as spatial domain, transform domain, and dictionary learning based according to the image representation. Using machine learning, sparse representations have become a trend and are used image and vision applications. The general idea of dictionary learning for image denoising by learning a large group of patches from an image dataset such that each patch in the estimated image can be expressed as a linear combination of only few patches from this redundant dictionary. The aim of the present paper is to demonstrate that both SVD and PCA has same task in image denoising provided that they are learned directly from the noisy image. In this paper, we present a result of comparison among four dictionary learning algorithms K-SVD, and local PCA, hierarchical PCA and global PCA applied on the Synthetic Aperture radar (SAR) despeckling task. The experimental results show that the proposed K-SVD algorithm is provide an adequate results in removing speckle noise from the SAR images. Keywords—Spares representation, dictionary learning, SAR images, Deseckling, K-SVD, PCA.