Image watermarking with optimum capacity

Abstract

Combining the embedding process with the requirements of the real watermark applications, a novel image watermarking with optimum capacity is presented in this paper and an implementation approach in Discrete Wavelet Transform (DWT) is given. An optimum capacity analysis is presented, forming theoretical basis for design of optimum systems. Watermark was generated randomly and embedded by modifying DWT coefficients of original image. In some applications, not the more information embedding is, the more performance is. The watermark embedded capacity is computed from the requirements of real watermarking applications, which are utilized false alarm probability and/or detection probability. Computation of detection probability and false alarm probability is given with both fixed and variable detection threshold. Robustness of this watermarking may be not the best, but the watermarking can meet requirements of real watermarking applications. Relationship between watermarking capacity and visual evaluation is deduced. Independent Component Analysis (ICA) is utilized to extract the watermark blindly. Experimental results have demonstrated that the proposed watermarking technique with optimum capacity can meet requirements of different real watermarking applications. It is indicated that this new watermarking technique has superior advantages over the existing ones in many aspects according to the real watermarking applications requirements.