Research On Edge And Color Information Based TV Logo Recognition

TV logo recognition is used to detect and recognize the logo of TV station from the input images. The TV logo recognition method is mainly related to technologies of image processing, image recognition, pattern recognition, etc. It is widely used in content-based video retrieval, program ratings survey statistics, radio spots illegal surveillance and other fields. In order to achieve the goal that the TV logo can be located quickly and recognized more accurately, a new TV logo recognition method based on edge and color information is proposed, which take full advantage of the intrinsic characteristics of TV logos such as position invariance, edge invariance and color invariance.Main contributions of the thesis are as follows:1) The overall designing of TV logo recognition algorithm is proposed. The main contents include developing the technology roadmap, designing the algorithms of detection, recognition and switching detection, and determining the key technologies and test methods.2) An improved TV logo detection algorithm based on time-averaged edges is proposed. This thesis utilizes priori knowledge-based approach to locate the area of the TV logo, and uses the edge information to extract the logo’s sample template image. The experimental results show that the improved TV logo detection algorithm can locate the logo more accurately, and extract the logo’s sample template image more completely. The location rate can reach more than98%.3) A hierarchical TV logo recognition algorithm based on edge and color information is proposed. The recognition algorithm includes a coarse recognition stage based on color feature and a fine recognition stage based on gray-scale information. In the coarse recognition stage, the one-dimensional color histogram feature vector obtained by HSV space is used to determine the set of logo candidates. In the fine recognition stage, the normalized cross-correlation is used to achieve matching and recognition. The experimental results show that the recognition rate of the hierarchical TV logo recognition algorithm can reach91%.4) A TV logo switching detection algorithm based on the zapping image and edge information is proposed. According to the characteristics of the TV logo switching, two cases are considered. One is to detect the zapping image, the other is to detect the edge information. The switching detection is completed respectively by detecting the occurrence of the zapping image and the change of the edge information. The experimental results show that the detection rate of the TV logo switching detection algorithm can reach96%.Theoretical analysis and comprehensive experimental results show that the proposed recognition algorithm can locate the TV logo region more quickly, recognize the TV logo more accurately, and detect the switching behavior more precisely. The overall recognition rate can reach more than83.79%.