A one-pass VBR video coding in fix size

The paper proposes a low complexity, one-pass, variable bit rate video coding algorithm which gets smooth quality and satisfies the constraint of constant storage size. According to the correlation among frames, it uses the long average-complexity of coded frames to predict the coding frame´s complexity. It allocates bits through the whole sequence according to storage size and the average complexity of each kind of frame and further it decides quantization scale. In most cases, the allocated bits differ from actually used ones. Thus it introduces a concept similar to virtual buffer fullness in TM5 as a feedback factor to tune allocated bits slightly. With feedback factors, it reaches a trade-off between used and remaining bits. Simulation shows that it tunes the quantization scale slightly and frequently, but, overall, the quantization scale changes slowly. It gets higher visual quality than TM5 and other one-pass VBR algorithms. It is very effective for constant storage size. Since it does not require used bits to be as close as possible to allocated ones, it contributes little to the CPU load and is suitable for real-time applications.