Dictionary learning has shown its great performances in image processing and computer vision. Modern learning tasks have to deal with large amount of data and thus put a high requirement on the efficiency of optimization algorithms. We propose a novel distributed algorithm for structured dictionary learning. Specifically, we leverage the existing alternating scheme with atom re-allocation followed by consensus updates. We also address to maintain the manifold structure of the dictionary atoms which is usually omitted by existing methods. Furthermore, advantages of the proposed algorithm are demonstrated for various applications.

This presentation is part of Minisymposium “MS49 - Image Restoration, Enhancement and Related Algorithms (4 parts)”
organized by: Weihong Guo (Case Western Reserve University) , Ke Chen (University of Liverpool) , Xue-Cheng Tai (Hong Kong Baptist University) , Guohui Song (Clarkson University) .

Authors:

Weihong Guo (Case Western Reserve University)

Yue Zhang (Case Western Reserve University)

Keywords:

computer vision, distributed dictionary learning, image reconstruction, machine learning