Abstract:
Text is the one of the compelling innovations of the society. It has portrayed an important
role in human vivacity since antediluvian times. As the text contains very important semantic
information that is used in diverse range of vision based applications, therefor nowadays
scene text detection is a very active research topic. Scene text detection is mainly about
detecting text within the surrounding environment, it could be indoor or outdoor.
Human beings inherently get the capability to perceive things within their environment however
it is a challenging task for computer systems to discover and acknowledge the textual
data in their environment. We humans are supposed to develop new technique to induce
artificial intelligence in them to make them work like humans. Because of the rapid development
in technology, numerous researchers have worked to generated wide number of
technique to detect and extract the text out of scene text images, but still there is a need of
cost effective and accurate scene text detection technique.
In current thesis, Scene text detection technique is proposed. Which is a combination of
two algorithms: (1) Image Rectification (2) MSER technique. Both algorithms perform
state of the art on test images and generate efficient results. Quantitative measures(with
state of art existing technique) are evaluated to prove the signicance of proposed algorithm.
Results reveal that the proposed technique is almost two times efficient and more accurate
comparatively state of art scene text detection technique.
