Sample Value Ordering for Audio Reversible Data Hiding

Pixel value ordering (PVO) was shown to be an efficient technique for high-fidelity reversible data hiding in images. This paper proposes an audio reversible data hiding scheme inspired by PVO, namely sample value ordering (SVO). Large uniform areas are uncommon in natural images, but sections of silence are relatively common in audio samples. In order to properly exploit the nature of the host audio file, a block-based approach is employed. Non-overlapping blocks are classified into three categories: smooth, somewhat complex and complex. All samples from a smooth block are considered as possible hosts using a classic histogram shifting based approach. For somewhat complex blocks, only the maximum and minimum sample values are used for data hiding with the PVO inspired SVO. Complex blocks remain unchanged.