Research On Text Location And Extraction In Color Images

As the rapid development of computer science and multimedia technology, the multimedia information, mainly composed of color images, has rapidly become an important general information media. Texts in color images usually contain much high-level semantic information. So how to locate and extract the text fast and accurately from color images has became the important research area in the world.For the text in color images, there exists distinct edge contour between the characters and the background. So locating and extracting the text from color images adopt edge-based method in the thesis. Color image enhancing is firstly handled and an effective vector median filtering algorithm modified is presented. The Prewitt edge detection operator is expanded in the color space, and is used in the edge extraction of color images. Subsequently, connected component is labeled, and the candidate for text region is located. Finally according to person eye's sight identity and text's characteristic property, the candidate text region is filtered to wipe off the false text region and extract text from images. For the sake of the statistics model's advantage embodied in pattern recognition research, this paper increases the correctness of text detection by expanding algorithm under support vector machines. Applying active learning support vector machines can reduce the number of examples effectively on the premise of keeping correctness of the classifier.