الفهرس 
Abstract
ConnectedComponentsAnalysisOnly 14 pages are availabe for public view
 |  | from | 118 |  |  |
| | | from | 118 | | |
Abstract
Understanding images have become a major goal of the computer vision field. Images contain many kinds of things (e.g. text, people, faces, and animals). Text is the most informative thing amongst the contents of the images. So, the fundamental goal is to detect and recognize text within the images.
There are many applications of scene text localization and recognition. Such applications are, helping visually impaired by automatically detecting any text in the advertising plates while their walking in the street and reading it out loud to them. Also, in the navigation where the scene text localization and recognition are needed to mainly transform textual information in a way to be easily processed. Automated translation to avoid the need for any manual input. Moreover, Indexing and searching image databases by textual content.
Since the problem of finding texts in natural images encounters many challenges, for example, the mixing of the texts with the background, which led to the decrease in the rate of texts retrieval and increase the number of false detections of texts, so we focus in this work on the issue of finding texts from scene images. This problem called Text Localization in Scene Images. Our main goal in the scope of this problem is to seek to raise the level of retrieval and reduce the false detectives, which is still difficult to overcome, especially in images with complex backgrounds.
After giving an introduction of the localization problem and the previous work of solving this problem in the first part of this thesis, the proposed methods which depending on the statistical inference model and some statistical measures are introduced. In additional, exploiting some geometric features and shape properties are used. It is shown that the proposed method is better than some exiting methods. Also, the proposed method has achieved the higher recall and precession especially in the complex images and the images that have blurring.