Graph recognition from image

Simon Birk (2010) Graph recognition from image. EngD thesis.

PDF Download (2027Kb)

Abstract

Optical recognition of documents constitutes an important part of computer vision. A document is usually presented as a sequence of images. This sequence is then processed by a document recognition system. During recognition process each image is separated into various elements. Each type of element requires a different approach to recognition procedure. We have developed an application that represents one of such approaches. The application recognizes a graph containing nodes and edges from an image. For each node location, size and label are recognized. For each edge location, direction and weight are recognized. Given the directions of edges the orientation of identified graph can be determined. As the result of recognition a structure is stored in the form of XML (eXtensible Markup Language), which is widely used and is therefore suitable for further processing by various algorithms.

Actions (login required)