In　this　paper,　a　new　robust　watermarking　scheme　is　described.　In　this　scheme,　two　complementary　watermarking　algorithms　are　used　to　resist　different　attacks.　First,　the　largest　singular　value　of　every　frame　is　calculated　after　discrete　wavelet　transform　(DWT)　and　segment.　After　single　value　decomposition　(SVD),　the　watermark　sequences　are　embedded　several　times,　for　redundancy,　by　adjusting　the　largest　singular　values　of　the　frames　according　to　the　pre-set　rule.　This　watermarking　algorithm　works　well　in　the　case　of　common　signal　processing.　The　second　watermarking　algorithm　based　on　the　histogram　is　utilized　to　resist　synchronization　attacks.　In　the　process　of　image　extraction,　two　types　of　watermarking　images　are　extracted　using　two　different　methods.　Then,　an　appropriate　judgment　is　enforced　to　determine　the　final　watermark　image.　In　this　way,　though　each　of　the　two　watermarking　algorithms　is　not　robust　enough　to　withstand　all　attacks,　they　have　complementary　advantages.　In　other　words,　by　embedding　two　types　of　watermarks,　the　whole　watermarking　scheme　displays　great　robustness　when　facing　both　common　signal　processing　and　synchronization　attacks.