Censorship of video images with adaptive reconstruction based on spectral domain watermarking

Censorship of video images has become an important field of science nowadays. Multimedia files often contain sensitive content that should be unseen by human observer. Therefor a novel system for censorship of video images based on spectral domain watermarking is described in the present paper. Faces are detected by face detection algorithms and embedded as watermarks into the images. Then these areas are blurred to protect privacy. The whole MJPEG video is also embedded as watermark to protect against tampering. Adaptive reconstruction allows low quality reconstruction of tampered areas and high quality reconstruction of protected faces.