Maximum likelihood sequence decoding of digital image watermarking

This paper explores possible methods to detect the watermark signal. We show how using sequence detection techniques, such as maximum likelihood sequence detection (MLSD), can provide the same performance as conventionally used correlation detection while simplifying the detection process. As a result, we can increase the dimension of the watermark signal (achieving higher capacity), and keep the increase in computational complexity manageable. We will also show that with MLSD, we can have a measure of confidence level of the detected signal that is directly dependent on the watermark signal to noise ratio (SNR_wm).