Improvement And Application Of Block-sparse Dictionary Learning Algorithm In Compressed Sensing

In compressed sensing,considering the structural information of the signal often leads to the case that they can be better expressed by a union of a small number of subspaces.Based on the sparse agglomerative clustering,the block structure dictionary learning method distinguishes the similarity with the intersection size of the dictionary atom's support sets,but it can not discern the difference in the size of the atom's support set.Aiming at this problem,proposed a method of learning block structure dictionary based on spherical K-means clustering,which divides the dictionary atoms' support sets into the metric,discriminates the similarity of the dictionary by the cosine distance,and form a dictionary with non-uniform block structure.Finally,we can use it to reconstruct the signal.The experimental results show that the matching between the dictionary and the image signal is better than that of the discrete cosine(DCT),unstructured dictionary and the block structure dictionary based on sparse agglomerative clustering,it can improve the image reconstruction quality effectively and reduce the signal reconstruction error.