An Automated HSV Based Text Tracking System from Complex Color Video

摘要

Tracking, extraction and recognition of text from video are important steps in building efficient indexing and retrieval systems for multimedia databases. We propose a top-down approach in which multiple cues are used for detecting text blocks in videos. In our proposed text tracking and extraction system, color and edge features are combined together to increase the precision of text detection. Our system detects both caption text as well as scene text of different font, size, color and intensity. Our objective is to detect both stationary text and moving text from complex color video. Extracted texts from videos are recognized using an OCR and stored in a database with relevant information. Such texts are used in future for retrieval of video clips based on any given keyword. This paper shows comparative result of RGB and HSV based approaches using different types of domain and shows the result of text tracking in different scenarios.