Research On Text Location In Complex Background Images

With the development of the multimedia technology, the study of locating texts under complicated background has not only enriched image processing theoretically, but also has enormous value in practical application, for example, the image retrieval under Internet environment and the discernment of the number plate in traffic administration. The location and extraction of text from complex background is an important research problem in the computer vision. The variation of the text in terms of characters font, size and style, orientation, alignment, texture, color and complex background makes the problem of text localization very difficult. The scene content is unconstrained and maybe both indoor and outdoor scenes under any lighting or contrast conditions.To extract complex background text, text areas should be find first which then subjected to recognize. Current text location methods are researched in this paper, and the advantage and disadvantage of them are analyzed. Then text location method based on edge detection and support vector machines is proposed.Edge detection based text location method is composed by Pyramid decomposition, edge detection based improved Canny algorithm, edge abstracting and binary, connected component analysis and text region identifying and combination. First, the improved Canny algorithm is used to detect the text edge, then connected component and text region identifying and combination is used to get the candidate text region. For the sake of the statistics model's advantage embodied in pattern recognition research, this paper uses the method of support vector machines classifier training to improve the correctness of text location. The active learning support vector machine is applied to reduce the number of examples effectively, and the result of the experiment is good.The result of the experiment shows that this algorithm can well and exactly locate the text, this algorithm is valuable in theory and application.