Abstract
In this thesis, we proposed a channel coding mechanism for unequal important data protection, such as video streaming or layered videos, based on the concept of Luby-Transform (LT) code in digital fountain. An Unequal important data is that different parts in data have different importance, and needs being protected according to their importance in different parts. Under network environments where clients have heterogeneous bandwidth and computing power, video streaming services that provide coarse-to-fine video quality are required because the usage environments are diverse and changing dynamically. The contributions of this thesis are that we analyzed the coding graph of LT codes to conclude the factors of decoding probability and redesigned the coding graph to achieve the unequal protection requirement. As compared to LT codes, such as Reed-Solomon code, the proposed channel coding approach can better performance on PSNR of video streaming and layered video coding. Furthermore, the decoding complexity of the proposed mechanism is very low so that it can be applied on the real-time decoders, such as mobile phones, which only have limited computing power and request low delay requirement. The experiment results present that the proposed approach reaches about 3 db PSNR improvements on other LT codes on average. More significantly, the results demonstrate that the proposed approach also has better performance when it is applied on layered videos. Finally, we also proposed an novel encoder architecture for the proposed codes, and the architecture provides a flexibility and convenience to design or generate all kinds of LT-like codes.