Automatic SystemforFiltering Obscene Video

Thispaperproposes anautomatic systemfor filtering obscene videosequences. At present, an enormous numberofvideo files areavailable onthevarious channels such asP2Pnetworks, spamemails, andwebsites. Especially recently, manyvideofiles arepropagated viaUCC community sites with themanualfiltering process toscreen theharmful videos. Thus, automatically recognizing obscene pictures invideo files becomes anextremely important problem. Inthispaper, wepropose a new,efficient, andfast algorithm fordetecting obscene videos. Ouralgorithm isbasedonsegmentation. We first extract the frames ofavideo, detect global motions, separate theforeground objects, andclassify eachframeasanobjectionable orabenign framebasedonskincolor, texture, andshapefeatures. Our algorithm showsa remarkable capability todetect obscene framesinthevideofiles ofvarious videoedits suchasUCC contents.