Text detection and its segmentation from TV screen

This paper presents a system for text extraction on images taken by grabbing the content from the TV screen. The main tendency is the image preparation for the OCR to improve its accuracy. Recognized text is used for automatic generation of TV menu system structure, which is used for verification of TV set operation. This system is used as a part of Black Box Testing system in order to test and functionally verify TV set operation. The improvement is tested on the range of the full image dimensions and the results are compared with those for the original image. The proposed method is robust for different text fonts and styles, and the results show a improvement in character recognition accuracy.