We have witnessed a big success of the network coding theory over the past decade, which assumes that relay nodes can encode their received data before passing them onto others in a multicast communication network, thus leading to an increased throughput, as the size of network-coded data is smaller than that of the original ones. In this thesis, a very unique scenario is considered, where such a relay node receives two bit-streams that are compressed versions of the same source image, coded by a DCT-based scheme at two different bit-rates. With the proposed schemes, it is demonstrated that much fewer bits are needed to be transmitted from this node without affecting the overall performance. The proposed schemes are based on asymmetric network coding, and exploit the intrinsic...[ Read more ]

We have witnessed a big success of the network coding theory over the past decade, which assumes that relay nodes can encode their received data before passing them onto others in a multicast communication network, thus leading to an increased throughput, as the size of network-coded data is smaller than that of the original ones. In this thesis, a very unique scenario is considered, where such a relay node receives two bit-streams that are compressed versions of the same source image, coded by a DCT-based scheme at two different bit-rates. With the proposed schemes, it is demonstrated that much fewer bits are needed to be transmitted from this node without affecting the overall performance. The proposed schemes are based on asymmetric network coding, and exploit the intrinsic relationship between the two compressed versions. More specifically, the network coding schemes consist of three steps: (1) entropy-decode each compressed bit-stream, (2) apply multi-level logic operations on those (quantized) DCT coefficients, and (3) entropy re-encode the result to form a new but fully compatible bit-stream. The proposed schemes mostly differ in step two, where four different and theoretically proven techniques are designed to make use of the underlying characteristics between the two streams, and simulations are also conducted to justify the usage of the techniques. Extensive experimental results show that substantial reduction can be achieved by our proposed network coding.