An Effective Rate Control Algorithm with Kalman Filtering and Two-Stage Rate-Distortion Modeling
| Autor: | Chih-Chien Dong, 董志堅 |
|---|---|
| Rok vydání: | 2001 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 89 This paper proposes an application of the historic Kalman filtering algorithm to the rate-buffer control problem for the H.263 standard. For effective handling of the tradeoff between temporal resolution and spatial quality, the proposed rate-buffer control framework uses a two-stage frameskip decision mechanism. Here the long-term decision is based on bit-rates and characteristics of windowed pictures, and the on-line decision depends on the estimator of the Kalman filter. The estimation model of the Kalman filter is to estimate the encoded bits of each frame for assigning frame target bits, and then use a prediction error feedback to stabilize the buffer fullness. Because a possible overflow is speculated via the Kalman filter estimating, the proposed rate-buffer control will drop the frames pre-selected in the first stage until the buffer falls into the safe state. While a frame is determined for exactly encoding, the proper picture-layer quantization parameter can be obtained from the measurement model of the Kalman filter. For the slice-layer modeling, an appropriate slice-layer rate-distortion model is designed and used to construct a constrained optimization problem for achieving a slice-layer bit-budget allocation criterion according to both the maximization of signal-to-noise ratio (PSNR) and the consistence of perceptual vision. After the rate controller uses this criterion and skillfully links the hierarchy relation between all kinds of quantization parameters, the macro-layer quantization parameters can be properly obtained for the practical quantization. In comparison with TMN8 model, the proposed rate control mechanism can simultaneously achieve better spatial and temporal performance. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|