Rate control for video coding using exchange of quantization noise and signal resolution
✍ Scribed by Hiroshi Ito; Yoshiko Hatano
- Publisher
- John Wiley and Sons
- Year
- 2000
- Tongue
- English
- Weight
- 981 KB
- Volume
- 83
- Category
- Article
- ISSN
- 8756-663X
No coin nor oath required. For personal study only.
✦ Synopsis
A new algorithm of rate control for video coding is proposed, which can be applied to any video coding schemes that contain transform coding such as DCT. The main feature of the algorithm is to jointly control the quantization parameter, Q, and the scanning length of coefficients, L, with L corresponding to the transmitting signal resolution. The overall coding artifacts will be decomposed into the quantization noise and the resolution noise, which are defined in this paper. The relationship of Q and L is discussed based on the asymptotic coding theory. A rate control strategy is derived from this discussion to stabilize the quantization noise. This is done by defining the maximum tolerable level of quantization noise and then exchanging the excessive quantization noise with a reduction of transmitting signal resolution. It is shown that aliasing artifacts due to controlling the length of coefficients are sufficiently suppressed by pre/postfilters. Simulation results are provided in comparison with the H.263 coding algorithm. When a test sequence of CIF size has been encoded in 512 kbps, a 4-dB reduction in quantization noise has been observed while the overall SNR has dropped by 0.9 dB. The main contributions of the proposed algorithm are (1) it produces subjectively pleasing reconstructed images by suppressing the quantization noise, (2) it is applicable for low-bit-rate video coding, and (3) it provides compatibility with existing coding standards such as H.263.