Extraction of Text Regions and Recognition of Characters from Video Inputs

In this paper, a new algorithm for extracting and recognizing characters from video, without a priori knowledge such as font, color, size of characters, is proposed. From input videos with complex backgrounds at low resolution, continuous frames with identical text region are automatically detected to compose an averaged frame. Using boundary pixels of a text region as seeds, we apply region filling to remove backgrounds from characters. Then color clustering is applied to remove remaining backgrounds. For the recognition of characters, simple features such as white run and zero-one transition from the center, are extracted. These features are compared with a pre-defined character feature set to recognize the characters. Experimental results tested on various news videos show that the proposed method is effective in terms of caption extraction rate and character recognition rate.