Robust and Accurate Method for Textual Information Extraction Over Video Frames