The process of locating a word as well as placing a bounding box around the word in a scene image is known as text localization. The degradations, which affect the scene image while capturing, increases the complexity of the task by many fold . Here, the text localization is carried out in video frames. The major factor that affects the video is motion. An algorithm has to be effective in countering the motion and locating the text in each frame. The work on text localization in videos is carried out by Vivek Veeriah, an intern from Coimbatore Institute of Technology, under the supervision of Prof. K R Ramakrishnan and Deepak Kumar. The details about the algorithm will be available on request.
The results of text localization from old algorithm at the beginning of the internship. There are totally eight videos split into small frames. You can watch all the videos on YouTube by clicking the link:
http://www.youtube.com/watch?v=WaGDEgk9o4I&feature=share&list=PLA_n7q-bw0T6itEFNNqievoKxxKywB8CS
The results of text localization at the end of the internship. Here, All the eight videos are combined and formed as a single video. Blue colored bounding boxes are the stable results obtained from the algorithm.