Abstract
dc:descriptionDocument image processing concerns the analysis and recognition of the text and it consists of several stages. Principal stage of the whole procedure is the binarization or thresholding stage, according to which the pixels are classified into text and background. It is a crucial stage that can affect further stages including the final character recognition stage. This thesis is focused on document image binarization, including both binarization techniques and evaluation methodologies. Particular focus was given to historical documents because of their difficulty that originates from the various degradations. Specifically, a performance evaluation methodology was developed that makes use of the skeleton of the characters at both the semi-automatic ground truth construction stage and the evaluation stage. Afterwards, the aforementioned methodology was improved and the ground truth images were more reliable containing lower subjectivity degree, while several different evaluation measures were studied during the development of the new measures. The new measures are based on (a) weights that start from the ground truth contour and (b) the local stroke width to limit the weights close to the character areas and to properly normalize those weights. Experimental results prove the validity and effectiveness of the new measures for document images, while other measures concern the image or signal processing area in general. As far as the binarization techniques are concerned, a new technique was developed as an improvement of an existing technique. The improvements offer better results for documents with fonts of various sizes and better faint character detection. To further enhance the quality of binarization and be more robust against different types of degradations, a new binarization technique was developed that was based on background estimation and on the combination of selected global and local binarization techniques. This technique was focused on handwritten documents and the experimental results show high performance for documents with various degradations such as faint characters, bleed-through and non-uniform background. Moreover, the aforementioned technique can be modified to be used on machine-printed documents with very promising results. Additionally, a binarization technique was developed for the binarization of the text areas captured from video content. This technique is based on an existing well-known binarization technique and sets low and high values in its global parameter for the inside and outside area of the text. Initially, the definition of the text areas is based on the baselines of the text and at the final stage the text areas are better defined by the convex hulls of neighboring textual components. Furthermore, through the document image binarization contests that we organized, a publicly available benchmark was created that aids in the development of document image binarization techniques and evaluation methodologies.
Degree
thesis:*- Grantor dc:publisher
- National and Kapodistrian University of Athens
- Year dc:date
- 2013
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Ntirogiannis, Konstantinos
- Ντιρογιάννης, Κωνσταντίνος
Subjects
dc:subject × 14- Προ-επεξεργασία
- Δυαδική μετατροπή
- Μετρικές αξιολόγησης
- Πρότυπη εικόνα αναφοράς
- Επεξεργασία ιστορικών εγγράφων
- Pre-processing
- Binarization
- Evaluation metrics
- Ground-truth image
- Historical document image processing
- Φυσικές Επιστήμες
- Επιστήμη Ηλεκτρονικών Υπολογιστών και Πληροφορική
- Natural Sciences
- Computer and Information Sciences
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/38182