Image Authentication and Restoration Using Irregular Sampling for Traffic Enforcement Applications

Traffic image sequences are important information for the purpose of traffic system control and traffic accidents surveillance, for example, the determination of offending vehicles. A semi-fragile watermarking method for the automatic authentication and restoration of traffic images using irregular sampling is described. Watermarks are embedded into the pinned field of the pinned sine transform (PST) of the original image, which reflects local malicious tampering on the texture of the image. When tampered blocks are detected, the restoration problem is formulated as an irregular sampling problem in approximation subspaces. These blocks are then reconstructed, making use of the information embedded in the same watermarked image, through iterative projections onto convex sets in approximation sub-spaces. The restoration process is a variant of the Papoulis-Gerchberg algorithm, and is robust to common image processing operations such as lossy transcoding and image filtering