Key Frame Extraction Algorithm For Educational Digital Video Data Using Text Property

摘要

PURPOSE: A method for extracting a representative screen using a subtitles characteristic in educating digital video data is provided to prevent a mal-detection and to create effective representation screen extracting information by reflecting a space variation of a figure in a screen properly and decreasing an excessively divided information properly. CONSTITUTION: In a scene conversion is detected by applying to learning video data, a scene conversion candidate area is extracted. More exact scene conversion time is extracted from the extracted scene conversion candidate area. A representative screen is extracted using finally extracted scene conversion information.