Abstract: In the large visual multimedia network threat interference noise, a lot of bad data and stable operation of the multimedia network,resulting in multimedia audio and video signal errors in visual and the sense of hearing, the performance is not good. A bad data of large visual multimedia network of linear superposition detail capture efficient detection method is proposed based on edge detection function, using Canny, get the multimedia network data exchange noise edge invariant moment threshold image segmentation, using empirical mode setting for bad data detection, feature vectors using nearest neighbor matching method can identify audio and video data on the potential of matching, using the standard segmentation method based on mathematical morphology to get the minimum cut criteria, realization of improved detection algorithm, the simulation results show that, the algorithm can effectively improve the bad data detection performance, strong anti interference performance and robustness.