Robust blind watermarking technique for color images using Online Sequential Extreme Learning Machine

In this paper, a robust blind color image watermarking technique using Online Sequential Extreme Learning Machine (OS-ELM) is proposed. Blue channel is utilized and transformed using DWT. Low frequency LL4 subband is used for watermark embedding. A variant of mini-batch machine learning algorithm i.e. OS-ELM is initially tuned with a fixed number of training data used in its initial phase and size of block of data learned by it in each step. The training data to OSELM is constructed by combining the quantized and desired LL4 sub-band coefficients of the DWT domain. A random key decides the starting watermark embedding position of the coefficients. Two binary images are used as watermark. The robustness towards common image processing attacks is enhanced using this process. Experimental results show that the extracted watermarks from watermarked and attacked images are similar to the original watermarks. Computed time spans for embedding and extraction are of the order of milliseconds, which is suitable for developing real time watermarking applications.