Application of histogram equalisation to audio watermarking
Finbarr B. O'Donovan · 2004
In previous work, the authors proposed the use of non-linear detection functions in spread-spectrum watermarking. When combined with side-informed embedding, these detection functions have yielded improved performance compared with linear detectors. The efficiency of these detectors is dependent on the statistical distribution of the extracted feature vectors. We now review these detectors and discuss a particular extraction function and its resultant feature vector statistics. We then show the need for, and the application of, histogram equalisation to these feature vectors. Finally, we present the results of applying these techniques to real audio signals.