Rate-Distortion Analysis for H.264/AVC Video Statistics [chapter]

Luis Teixeira
2011 Recent Advances on Video Coding  
A video coding algorithm focus on the trade-off between the distortion and bit rate, where usually to a decreasing distortion corresponds an increasing rate and vice-versa. In R-D theory, the R-D function allows to estimate the lower bound for the rate at a given distortion. However, this value may not be possible to obtain in practical video encoders implementations. Operational R-D (ORD) theory applies to lossy data compression with finite number of possible R-D pairs (Schuster & Katsaggelos,
more » ... 1997a ). Rate Operational Rate-Distortion Rate Distortion Model Distortion www.intechopen.com Rate-Distortion Analysis for H.264/AVC Video Statistics 119 Some rate control schemes incorporate spatio-temporal correlations to improve the accuracy of R-D models, by using statistical regress analysis for dynamical model parameters update. Representative of this approach is the MPEG-4 Q2 (Chiang & Zhang, 1997) , and the linear MAD models (Lee et al., 2000) , where model parameters are updated by linear regression method from previous coded parameters. H.264/AVC JM rate-control algorithm also uses a quadratic rate model. In addition, the H.264/AVC rate-control solves "chicken-and-egg" dilemma as the Lagrange multiplier is modelled as a function of quantization parameter (Wiegand & Girod, 2001) . Rate-quantization relationship can be used to compute the quantization parameter. Nevertheless, the model-based rate functions frequently depend on the complexity of the coding unit that is obtained after the rate-constrained motion estimation and mode decision with the Lagrange multiplier. The JM algorithm of H.264/AVC proposes a linear prediction model to solve this problem by estimating the mean of absolute difference (MAD) from the previously coded units. Then the quadratic model can estimate the quantization parameter. However, rate-distortion re-analysis can be further investigated based on the coding characteristics of the H.264/AVC for improving the coding performance (Kamaci et al., 2005; Ma et al., 2005) particularly in the case of joint video coding and the use of different distortion metrics. We may find in the literature extensive studies regarding optimizing a video encoder encoder with R-D considerations include mode decision (In summary, to optimize a video encoder, the rate-distortion optimization techniques play a very important role. R-D models are functions that predict the expected distortion at a given bit rate. This is very important for joint video coding applications that attempt to optimized quality, e.g. minimize distortion, in environments where the channel conditions vary dynamically or the number of broadcast programs varies through time. Thus in this section we propose to present and evaluate several R-D models. At the same time, we propose also to study the bit rate variability as a function of the video quality (Seeling et al., 2004 (Seeling et al., , 2007 . This type of analyse is typical of a communication network perspective. By re-analyzing the characteristics of the bit-rate and the data in the transform domain, a simple rate estimation function can be obtained that will allow support the allocating of video bandwidth within different video programmes. Rate control in international standards Although the MPEG video coding standard recommended a general coding methodology and syntax for the creation of a legitimate MPEG bitstream, there are many areas of research left open regarding how to generate high-quality MPEG bitstreams. This allows the designers of MPEG encoder great flexibility in developing and implementing their own MPEG specific algorithms. To optimise the performance-of an MPEG encoder system, it is important to study research areas such as motion estimation, coding mode decisions, and rate control. The main goal of rate control is to manage the process of bit allocation within a video sequence and thus the quality of the encoded bitstream. Regarding rate control, encoders www.intechopen.com Recent Advances on Video Coding
doi:10.5772/16945 fatcat:evlyn4o4kfddpewbcz4szvo3oa