Captions in news video frames contain rich information of semantics. It is helpful to the content-based news video retrieval if we can detect and locate these information. In this paper, we study the location features and types of news captions, and apply processing technology, including image analysis, image segmentation, image enhancement, in order to find a way to deal with caption detection and localization in news video frames.In this paper, we first study the typical methods of text detection and text localization in video frames, and analyze types and features of captions in news video frames. Then, we provide a method of caption detection based on wavelet transform and a method of caption localization based on edge detection. At last, we design and implement the caption detection module and caption localization module, which belong to the system of caption extraction in news video frames. Experiment results further verify the feasibility and validity of the algorithm. |