5 - Computer vision applications

Abstract

This chapter provides a comprehensive coverage of applications of both directed and undirected PGMs to major computer vision tasks. It starts with a review of different computer vision tasks and classifying them into three categories: low level vision tasks, middle level vision tasks, and high level vision tasks. It then provides a detailed review of existing studies that involve PGMs for tasks at each level. Specifically, for low level vision tasks, this chapter discusses the applications of both MRFs, CRFs, and BNs for such tasks as image labeling, image segmentation, and image denoising. For the middle level vision tasks, this chapter covers the applications of both directed and undirected PGMs to such middle level vision tasks as object detection and recognition, object tracking, scene recognition, and 3D reconstruction. Finally, for high level vision tasks, this chapter discusses applications of different types of PGMs to various high level vision tasks, including facial expression recognition, human gesture recognition, human action and event recognition, and complex human activity recognition.