Daniel J. Louw and Haruhiko Kaneko
Video Coding, Image Compression, Side Information
Distributed video coding (DVC) is a recent coding paradigm that allows for the computational complexity of the system to be shifted from the encoder to the decoder. This makes DVC codes perfect for use in limited power wireless devices as well for distributed systems in the multi-view case. In this paper, a single-view multi-hypothesis transform domain DVC codec, which uses non-binary quasi-cyclic LDPC codes, is developed. A scaling factor for improving the correlation noise modeling of individual hypotheses as well as a new method for combining these hypotheses is presented. A closed form solution for optimal reconstruction based on the new method is derived. The simulations show that the system is robust and able to effectively compress a variety of sources. The system shows gains of up to 0.6 dB over state of the art DVC systems and gains of 2 dB over the intra mode of the h.264 codec.
Important Links:
Go Back
