Anzeige
Erschienen in:
2018 | OriginalPaper | Buchkapitel
Recognition System to Separate Text Graphics from Indian Newspaper
verfasst von : Shantanu Jana, Nibaran Das, Ram Sarkar, Mita Nasipuri
Erschienen in: Operations Research and Optimization
Verlag: Springer Singapore
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Abstract
Identification of graphics from newspaper pages and then their separation from text is a challenging task. Very few works have been reported in this field. In general, newspapers are printed in low quality papers which have a tendency to change color with time. This color change generates noise that adds with time to the document. In this work we have chosen several features to distinguish graphics from text as well as tried to reduce the noise. At first minimum bounding box around each object has been identified by connected component analysis of binary image. Each object was cropped thereafter and passed through geometric feature extraction system. Then we have done two different frequency analysis of each object. Thus we have collected both spatial and frequency domain features from objects which are used for training and testing purpose using different classifiers. We have applied the techniques on Indian newspapers written in roman script and got satisfactory results over that.