Reconstruction of missing blocks in image transmission by using self-embedding

A new model of the content reconstruction problem in self-embedding systems, based on an erasure communication channel. In addition to the content hashes for authentication purposes, an encoder embeds in the image a reconstruction reference, which describes the content, and which can be used by a dedicated decoder to restore the tampered image fragments. Hence the term self-embedding, coined in the original publication. The proposed method is based on an alternative approach to spreading the reference information over the whole image, which has recently been shown to be of critical importance in the application at hand. Our paper presents a theoretical analysis of the inherent restoration trade-offs. A number of content reconstruction schemes have been proposed so far with various applications in mind, ranging from content authentication   through error concealment to privacy protection . Regardless of the application at hand, the reference information needs to be communicated to the decoder through an unreliable channel i.e., the tampered digital image. The average reconstruction quality, measured on 10 000 natural images is 37 dB, and is achievable even when 50% of the image area becomes tampered.