Font Size: a A A

Research On Channel Logo Recognition And Retrieval

Posted on:2014-06-16Degree:MasterType:Thesis
Country:ChinaCandidate:J H ZhangFull Text:PDF
GTID:2268330422463734Subject:Systems analysis and integration
Abstract/Summary:PDF Full Text Request
With the development of radio and television broadcasting technology, radio andtelevision have already penetrated every aspect of people’s daily life and work, andtherefore the demand for radio technology have arisen. Considering the important role ofchannel logo in distinguishing TV channels, it is very necessary to realize the computerautomatic identification of channel logos. Firstly, the channel logo is a distinct mark todifferentiate itself from other channel logos, which can be also used to protect itscommercial interest. Secondly, channel logos detection can be used to detect cablepay-per-view channels rapidly and help image processing tools to remove logos andimprove video quality. Thirdly, since cable TV via the satellite network has developed as aprogramming vehicle, the tracking of a channel logo can be employed to monitor thesignal status of a particular TV channel at the local transmitter side to secure safebroadcasting. Therefore, we need to choose an appropriate algorithm to recognize channellogos.Firstly, this paper completed the channel logo recognition and retrieval system basedon the content-based image retrieval technology. The method used image local featuredescriptor algorithm to extract visual feature vector of channel logo images, then introducethe vector quantization to produce visual word from the visual vector, utilize invertedindex and word frequency vector technology in text retrieval to build a fast index structure.The method used to data query is based on the approximate nearest neighbor search byresidual vector quantization.Secondly, this paper proposed channel logo retrieval method which is based on thefeatures of oriented gradients histogram and methods to support vector machineclassification. This method extract the features of channel logo image direction gradienthistogram, then use support vector machine method to classify feature data. Finally, thismethod realized the recognition of channel logo.The experimental results show that the two channel logo retrieval methods weproposed here can retrieve channel logos accurately. Logo retrieval system based on imageretrieval method increase the accuracy of detection, and will not be influenced by thebrightness, noise or background. In conclusion, this system has high practical value andcan be applied to video retrieval system. Similarly, logo retrieval system based on histogram of oriented gradients and support vector machine shorten the recognition timeand improve the recognition efficiency, more importantly, this system is user-friendly.
Keywords/Search Tags:Channel Logo, Content-based image retrieval, Histogram of orientedgradients, Support Vector Machine
PDF Full Text Request
Related items