Text detection is a challenging task when it comes to natural scene images (think of windows or other architectural patterns which have similar properties as printed text). Moreover, if the degree of freedom is increased and rotated text needs to be detected too, simple line searching algorithms do not suffice anymore.
This practical work will support the READ project which is a EU granted project dedicated to mass digitization of medieval documents from archives and libraries. Within this project, the CVL will develop document analysis methodologies such as form recognition or layout analysis.
A text detection algorithm will be implemented in this practical work. The algorithm needs to be able to correctly locate text of arbitrary orientation, font and size. The student will learn to implement state-of-the-art Computer Vision and Machine Learning algorithms.
- Matlab or C++ knowledge
- Machine Learning/Computer Vision knowledge