The sparse based representation plays a vital role in image processing, which represents a signal by linear combination of some atoms from the pre-defined dictionary. The atoms of the dictionary are either selected using k-means Single valued decomposition algorithm k-SVD or from any available transform. The process of deriving the best solution among the set of sparse representation is called sparse coding. The various sparse coding methods are categorized as greedy method and constrained optimization strategy. The Matching pursuits (MP), orthogonal matching pursuits (OMP) and their variants are based on greedy methods. Also the Gradient Projection, Interior point method, ADMM and their variants are based on constrained optimization strategy. Different applications of sparse based representation are Image reconstruction, Image Inpainting, Image denoising which uses dictionary learning approaches.
Volume 11 | 07-Special Issue
Pages: 1407-1413