Video querying via compact descriptors of visually salient objects | IEEE Conference Publication | IEEE Xplore