Special issue on Graph-Based Representations in Computer Vision

摘要

Graph-based representations are of pivotal importance in computer vision, pattern recognition and machine learning. They arise when the objects to be identified are decomposed into parts and relationships between them. Such representations are quite natural and find applications in low level image processing, such as segmentation or image filtering, and in high level vision tasks such as pattern matching. Graph representations also pose unique problems in machine learning, since they are non-vectorial in nature and require new methodology to be developed if they are to be learned from image data.