Rate-distortion modeling for scalable video coding

Scalable video coding allows for post-encoding adaptation of the coded bit stream. An appropriate rate-distortion model of the embedded bit stream can be useful for deriving closed-form relations between the parameters of the scalable encoder and the time-varying channel condition values. With a well-defined model, an adaptation mechanism can quickly update the optimal values of the encoder parameters within each transmission period, and hence, a real-time implementation may be realized. In this paper, two simple and accurate empirical (hyperbolic and exponential) models are developed for modeling the rate-distortion characteristics of a scalable video encoder based on motion-compensated temporal filtering (MCTF) and two-dimensional (spatial) discrete wavelet transform (2D-DWT). The experimental results demonstrate that the hyperbolic model has an acceptable approximation error (0.2~0.85dB) relative to the actual measurements, however, the exponential model demonstrates more error (0.75~3dB) relative to the actual measurements.