The Study Of Locating And Abstracting The Text Under The Complicated Background

The study of locating and abstracting texts under complicated background is of important to such an information age of today. It has enriched image processing theoretically. It has enormous commercial value in practical application, for example, the search of the image and video under Internet environment and the discernment of the number plate in modern traffic administration. So how to locate and draw the text fast and accurately from the image and video under complicated background has became the important research area in the world.The thesis analyses the segmentation of texts, text features and their recognition. We introduce and analyze several methods of locating and abstracting texts under complicated background.We perform the three main corner-detected methods: Harris, IPAN and SUSAN, and compare the quality between them. We improve the SUSAN corner-detected method, on the foundation of analyzing the method. We can find the better effect from the experiment. We present the algorithm of locating and drawing text on the SUSAN corner-detected, and expand it to the support vector machines frame.We introduce the Vatriant Gray Histogram (VGH) concept, and use it to locate and draw the text in succeed. We perform the algorithm of locating and drawing the text on the VGH, this method gains the perfect effect for the text/caption in the image and video. We integrate it with Fuzzy Theory in order to improve the robustness, preciseness and speed of this algorithm.Finally, we discuss the text detected in the video.