A Comparison on Sparse Coding and Moran's I Method for Image Denoising

Abstract

Image denoising is crucial to improve the quality of image visual, their effects, and/ or facilitating image analysis and processing. Image noise can appear in many imaging applications such as remote sensing surveillance and assistant of medical surgery. Noises are often introduced during the image acquisition process when the image acquisition sensor is being interfered. Hence, the image denoising technique is commonly used to restore the original signal through the estimation and approximation. Recently, a sparse coding technique employing the dictionary learning method has been used for image denoising. In this study, we compare a recently proposed image denoising method called Moran's I Vector Median Filter (MIVMF) with the sparse coding method and a traditional scalar median filter for the impulse noise. In these preliminary results, the sparse coding does not give satisfactory results as what we expected. Instead, the MIVMF has the best denoising results.