K-SVD Dictionary Incremental Learning Algorithm Based on Clustering Theory

To solve the time-consuming problem of K-SVD, a dictionary learning algorithm based on clustering theory is proposed. The proposed algorithm learns incremental dictionary from an incremental training sample set, then leverages clustering theory to combine the initial dictionary and incremental dictionary. Thus it provides a two way optimization for the entire training sample set and significantly increase the efficiency of the K-SVD dictionary learning algorithm for big sample set learning. The image super resolution reconstruction experiment demonstrates that the proposed algorithm exhibits the capability for incremental learning.