TV Logo Detection And Recognition Technology Research

It attracts more and more attentions on protection of legal broadcasting station withthe development of modern digital information technology, especially in the field ofdigital television technology. It is known for us that there is a logo in the picture oftelevision signal which has several advantages of vivid, easy-distinguished, stabilization,particular and so on. Logo is not only one of the eye’s distinguishing system, but alsothe important sign between medias. So logo becomes the reliable evidence of detectingunknown video.If we use the manual monitoring way, it often makes mistakes in the process ofmonitoring, it has not only low efficiency, but it is also easy influenced by theenvironment. The most important significance of logo matching research is that weshould design an effective auto distinguishing technique in the series of videos, whichafford the foundation of subsequent video analysis, understanding, and reference and soon. The logo distinguishing system consists of the following parts: logo segmentation,feature description and logo matching. The main contributions and novelties of ourwork are as follows:Firstly, an important method is proposed to improve the effect of logo segmentationcompared with the classical method which based on image gray threshold. We get thearea of logo though this method, then we find the edge of the logo by using the methodof Canny, thus we can use the predigested morphological watershed method to fill in thelogo area, as the result we get more pure logo.Secondly, after comparing several method of corner detection, Harris arithmetic,Susan arithmetic, DoG arithmetic and so on, because Susan arithmetic has a feature ofno gradient operation, so we take the improved Susan method to find the corners of thelogo based on the binary image.Thirdly, because of the low resolution and complex background, several classicalmatching methods are limited in the use of logo matching. After considering the factorof logo image deformation, the logo is matched after working out the matching cornersin transformed polar coordinates based on pairs of corners.All above algorithm are simulated on the platform of Visual C++6.0and Matlab7.0with several local TV videos; the experiments’ results indicate that this detectionand recognition algorithm proposed is effective and satisfactory.