Multimedia data recovery using information hiding

We study certain aspect of data hiding problem, not for the classical application of copyright protection, but instead for media data recovery. A medium data, for example, an image, is partitioned into N components. A content-associative signature of each component is generated and inserted imperceptibly into remote region of the image. When imperfect transmission of image causes loss of blocks, the embedded corresponding signature is extracted and used to recover the lost blocks of data. Compared to conventional approaches, this new method can provide the knowledge of the lost block content, in addition to neighborhood information, to the decoder at the receiver end. With the knowledge of the lost block content, better restoration results can be reached, especially when conventional neighborhood interpolation method is used together with the data hiding approach.