An Instructive Case Study with Face Recognition

• Published in: Dictionary Learning in Visual Computing pp. 103 - 108
• Main Authors: Li, Baoxin, Zhang, Qiang
• Format: Book Chapter
• Language: English
• Published: Switzerland Springer 2015; Springer International Publishing
• Series: Synthesis Lectures on Image, Video, and Multimedia Processing
• ISBN: 9783031011252; 3031011252
• ISSN: 1559-8136; 1559-8144
• DOI: 10.1007/978-3-031-02253-1_5

Sparse representation based on data-adaptive dictionary learning has evolved into a powerful, computational scheme, with different algorithms developed for various types of applications. The examples and discussion in the previous chapters only serve to illustrate a fraction of the diversity and general applicability of sparse representation and its various algorithms. In this chapter, we use a well-studied application, face recognition, as a case study to illustrate the general design strategies in applying the dictionary-learning-based sparse techniques. This is largely based on the approaches reported in Wright et al. [2009c] and Zhang and Li [2010a]. Such an instructive case study will serve to demonstrate the following typical tasks one would often need to address in building a dictionary-learning-based solution to a given problem: coming up with a proper dictionary-based formulation, finding a solution (or an approximate solution) for the learning task under the formulation, understanding the behavior of the the solution (e.g., convergence analysis of an optimization algorithm), developing inference schemes such as classification (if needed) under the learned dictionary, and extending a basic model by incorporating additional constraints specific to a given problem, etc.