To enhance image quality and retain a limited bit budget, intra-frame in the frame type selection plays a significant role in video coding systems because the frame is treated as a key frame for temporal domain reference. This paper presents an algorithm to predict accurate quantization parameters by developing variance-based distortion measurements. Firstly, the intra-frame rate control in JVT-G012 is analyzed before discussing possible drawbacks. Current algorithms includes JVT-G012 and Yan's algorithm based on mean square error metrics for intra-frame rate controls are compared to show that the proposed distortion metric is useful to estimate quantization parameters. Experimental results show that the proposed algorithm can significantly improve video quality up to 0.19 dB and 1.28 dB, compared with the algorithm JVT-G012, which is implemented in the H.264 reference software JM 15.0 and Yan's algorithm.
- Quantization parameter
- Rate control