Hybrid models of the rate-distortion characteristics for MPEG video coding

The rate-distortion based rate control scheme plays an important role in various video/image coding methods. How to acquire the rate-distortion characteristics of the image data in MPEG video coding is addressed in this paper. The proposed exponential function based rate-distortion model for an intra-coded frame (I frame) employs three control points to estimate a complete R-D curve for MPEG video coding. The performance of our method is comparable to that of Liang-jin Lin and A. Ortega (see IEEE Trans. Circuits and Systems for Video Technology, vol.8, no.4, p.458-46, 1998) or better than that of Tihao Chiang and Ya-Qin Zhang (see IEEE Trans. Circuits and Systems for Video Technology, vol.7, no.1, p.246-50, 1991). Based on the assumption that the error signal in a motion-compensated frame is a Laplacian distribution, another model is utilized to estimate the coding rate for P and B frames. Our hybrid model could be used in bit allocation for real-time video encoding due to its computational complexity and estimation accuracy