A Novel Blind Digital Watermarking Scheme Based on SVD and Online Sequential Extreme Learing Machine

Neelam Dabas, Geetanjali Kher, RamPal Singh, Vikash Chaudhary


With the development of internet, it is very easy to distribute and access the digital media. But with the distribution, security of intellectual property becomes a critical issue. Watermarking is one of the primary tool to protect intellectual property. A novel digital watermarking scheme has been proposed by using SVD and OSELM in IWT domain. This proposed scheme is divided into three phases. During the first phase of the embedding process, the cover image is transformed through Integer wavelet transformation (IWT) to get the low and high energy coefficient. The LL sub-band is divided into non-overlapping coefficients blocks and applied to SVD to get the singular values. Then watermark is embedded to modify these singular values. Second phase is the training phase in which an Online Sequential Extreme Learning Machine(OSELM) is trained to learn the relationship between the original coefficient and their watermarked version. The third and the final phase is the extraction phase in which watermark is extracted from the image by using the trained OSELM. This proposed method is the blind method as no original cover image is required in the extraction phase. Experiments have been performed on various images by attacking them with noise, cropping, blurring etc. Peak Signal to Noise Ratio(PSNR) and Bit Error Rate(BER) are calculated. Less BER value show that extracted watermark logo is very much similar to the original one and works efficiently to prove ownership.


Full Text:


DOI: https://doi.org/10.26483/ijarcs.v8i3.3187


  • There are currently no refbacks.

Copyright (c) 2017 International Journal of Advanced Research in Computer Science