Computer Science and Management Studies

Recognition of text in natural scene pictures is changing into a distinguished analysis space owing to the widespread availability of imaging devices in low-priced client product like mobile phones. Detecting text in natural pictures, as hostile scans of written pages, faxes and business cards, is a crucial step for variety of laptop Vision applications, like processed aid for visually impaired and robotic navigation in urban environments. Retrieving texts in each indoor and outdoor environment provides discourse clues for a good kind of vision tasks. During this project, we execute two processes like text detection and text recognition. In text detection, utilize contrast map is then binaries by median filter and combined with Canny's edge map to spot the text stroke edge pixels supported feature extraction. The options extractors are Harris- Corner, Maximal Stable Extremal Regions (MSER), and dense sampling and Histogram of Oriented Gradients (HOG) descriptors. Then implement text recognition. The primary one is coaching a character recognizer to predict the class of a character in a picture patch. The other is coaching a binary character category for every character class to predict the existence of this class in a picture patch. The two schemes are suitable with two promising applications associated with scene text that are text understanding and text retrieval. In additional we tend to extend this idea with word level recognition with lexicon methods with correct results. And additionally recognition text in real time pictures, videos and mobile application pictures.