Optical Character Recognition for scene text detection, mining and recognition

Text detection in scene images is an important prerequisite for many content-based image analysis tasks. In this method is mainly to identify an accurate and robust method for detecting texts in scene images. A fast and effectual cropping algorithm is designed to extract multi oriented text from an image. The input image is first filtered with connected component approach. Connected component clustering is then used to identify candidate text regions based on the maximum difference. The frame of each connected component helps to separate the different text strings from each other. Then normalize candidate word regions and determine whether each region contains text or not. The scale, skew, and color of each candidate can be estimated from CCs, to develop a text/non text classifier for normalized images. In this techniques not only detect text, it also extracts from the image and recognizes the text in terms of storing the recognized words into a separate file by incorporating several key improvements over traditional existing methods to propose a novel CC clustering based scene text detection method, which finally leads to significant performance improvement over the other competitive methods.