An　improved　video　retrieval　algorithm　based　on　dynamic　time　warping　was　proposed　to　solve　the　practical　problems　of　the　original　video　retrieval　algorithm　based　on　video　fingerprints,　such　as　offset　of　color　or　brightness,　abrupt　intense　disturbance　and　video　re-editing.　Video　frames　were　divided　into　segments,　then　the　new　algorithm　compared　the　change　of　video　fingerprints　instead　of　comparing　video　fingerprints　directly　and　used　a　modified　dynamic　time　warping　algorithm　to　cope　with　video　re-editing.　To　evaluate　the　performance　of　the　new　algorithm,　a　large　video　dataset　which　consists　of　74　TV　programs　and　154　re-edited　programs　was　set　up.　Experiment　results　demonstrated　the　effectiveness　and　robustness　of　the　algorithm.