We develop a -domain characteristic-based rate model for video transmission. Specifically, we extract three characteristics from the quantized discrete cosine transform spectra to construct an accurate rate-quantization (R-Q) model. Moreover, the extraction of the proposed characteristics requires only simple additive operations (additions, subtractions, and comparisons) and few multiplications. Based on the -domain characteristic-based bitrate model, we develop an efficient two-level greedy-based rate controller for low delay video transmission. Extensive experimental results show that due to the efficiency and accurate estimation of both macroblock and frame-level R-Q models, the proposed rate controller better maintains the source bitrate to meet a predefined buffer delay time, and substantially enhances visual quality as compared with other rate controllers.