Image Watermarking Based on Modified Harmonic Mean Filter

In this paper, a spatial domain image watermarking method based on the modified harmonic mean filter is proposed. In the embedding process, the blue component of a color image in the RGB color space is modified to embed a binary watermark image with the same size. In the extraction process, the embedded watermark is blindly extracted without the need of original image. The harmonic mean filter is considered and modified by using the concept of outliers in the filtering area. The modified harmonic mean filter is then applied to the watermarked image to obtain a better prediction of original image. The extracted watermark can finally be achieved by subtracting the predicted image from the watermarked one. Set of experiments are carried out to verify its efficiency. The performance of the proposed method in terms of weighted Peak Signal to Noise Ratio and Normal Correlation is evaluated and compared with the previous watermarking methods. Its robustness against attacks is also evaluated using the Stirmark benchmark. According to the results obtained, the proposed method provides a higher robustness on average at equivalent quality of watermarked image, compared to the existing methods.