作者: Ke Ren Wang , Wen Xiang Li
DOI: 10.4028/WWW.SCIENTIFIC.NET/AMR.998-999.1138
关键词:
摘要: Video steganalysis takes effect when videos corrupted by the target steganography method are available. Nevertheless, classical classifiers deteriorate in opposite case. This paper presents a to cope with problem of mismatch for detection motion vector (MV) based steganography. Firstly, Adding-or-Subtracting-One (AoSO) feature against MV and Transfer Component Analysis (TCA) domain adaptation revisited. Distributions AoSO various methods illustrated, followed potential TCA feature. Finally, experiments carried out on cases mismatch. Performance results demonstrate that TCA+AoSO significantly outperforms feature, is more favorable real-world applications.