Adaptive data embedding using VQ and clustering

Recently, embedding large amounts of secret data into an image with low distortion has become an important research issue in data hiding schemes, even though reversibility cannot be achieved. Because Shie et al. 's data hiding scheme adoptively adjusts thresholds to offer a huge hiding capacity at the cost of large and uncontrolled image distortion, we apply VQ and clustering techniques to achieve the same objective while maintaining stable and acceptable image quality of the stego-images. Experimental results confirm that our proposed scheme provides the same hiding capacity as done by Shie et al. does, and has better performance in image quality of the stego-image. Furthermore, on average, the PSNR of our proposed scheme is about 3 dB larger than with either Du and Hsu's or Shie et al. 's scheme.