Digital watermarking and tree structure based image quality estimation

A new semi-fragile-watermarking-based quality estimation scheme is proposed in this paper. The watermark is embedded into the tree structure of an image with proper embedding strength decided by statistically analyzing the characteristics of the image. In the watermark embedding, the bit-plane coding mechanism is used to simulate the general distortion that images will undergo. A universal "Ideal Mapping Curve" is generated using the SPIHT (Set Partitioning in Hierarchical Trees) algorithm with different BPP (Bits Per Pixel) settings. In the watermark extraction, the quality of the watermarked images will be estimated under different distortions, such as JPEG compression and Gaussian noise pollution. The "Ideal Mapping Curve" used for quality estimation against different distortions can be achieved by weighting the universal "Ideal Mapping Curve" using the empirical weighting factors. The experimental results suggest that the proposed algorithm works effectively and efficiently.