Blind Compressed Image Watermarking for Noisy Communication Channels
MIT Lincoln Laboratory Lexington United States
Pagination or Media Count:
Visually-distorted images can contain valuable information. Indeed, in tactical MANET networks where throughput is extremely valuable and difficult to come by, guaranteeing the delivery of every packet of an encoded image is impractical.However, designing a watermark that is resilient to the types of visual distortion imparted on an image or video due to channel losses is a difficult task. In this work, we introduce a new watermarking scheme for JPEG-compressed images that incorporates ideas from compressed sensing CS to achieve robustness against certain types of errors induced by noisy communication channels. This work uses CS techniques to embed a sparse watermark into L randomly selected quantized JPEG image coefficients. Sparse reconstruction techniques are then used to reconstruct the watermark from the coefficients that were received, including those that were incorrectly decoded. Through the development of this watermarking scheme, we would like to demonstrate and explore the effect of the error resilience properties of CS encoded signals on the image watermarking problem. We show through simulation that even with significant visual distortion in the received image, the CS encoded watermark can be detected with very high probability.
- Radio Communications
- Computer Systems