A novel statistical distortion model of mpeg-4 FGS
Xie Li, Wenjun Zhang · Proceedings of the IEEE Symposium on Emerging Technologies, 2005. · 2005
Laplacian distribution has been widely used to characterize the distribution of DCT coefficients. But in our research, we found that there are big discrepancy between the real data and the distortion model based on Laplacian distribution. And what's more, the higher the rate is, the larger the distortion is. In this paper, we proposed a new distortion model based on mixed Laplacian and uniform distributions. Experiments' results show that the MLQ distortion model can approximate well the real empirical data and have a higher degree of accuracy than conventional distortion model based on Laplacian distribution. Compared with the conventional distortion model based on Laplacian distribution, the new model can decrease the MSE distortion by 2~6 units. MLQ distortion model can represent the relationship of rate and distortion of mpeg-4 FGS enhancement layer with a higher degree of accuracy