Automatic System for Text Detection and Localization Using Cellular Automata

The cellular automata are a discrete and abstract way of representing dynamic model which change state based on some relationship with other members in the system. These patterns are used to map an object within the image. The present research used the cellular automata-based system to detect the text region in the natural seen image. Automatic text detection is gaining attention day by day due to its versatile range of applications. The present research used cellular automata to develop a system to detect Gurmukhi text in the natural scene images. The natural scene may contain signboard images, text on banners, text written on walls and text viable on any vehicle. The Gurmukhi script has its own set of unique features which helps to classifier to recognize. The efficiency of the text recognition heavily depends on the text extracted from the images. An algorithm was developed to detect and localize the text (in the present study, only Gurmukhi script is considered for study) in the natural scene images. In absence of benchmark dataset of natural scene images, we have developed our own dataset to test the efficiency of the system. The system was tested with well-known matrices named as recall metric, precision metric and P-value.