Rate Control in Video Coding
Zongze Wu, Sheng-Li Xie, Kexin Zhang, Rong Wu · InTech eBooks · 2011
Recent Advances in Video Coding 80 1.2 History of rate control In recent years, rate control has been the research focus in the field of video coding, many scholars and experts have achieved a lot of research achievements in the video rate control.The rate control in the video coding was proposed in 1992.The core of TM5 rate control algorithm is, under the situation that buffer is not overflow or overflow, distributing bits and determining the reference value of quantitative parameter by estimating the global complexity of the encoding frame, and adjusting the quantitative parameter by the activity of each block.In 1997, Chen (Chem., Hang.H.M., 1997) proposed a rate control algorithm which adjusting the frame rate adaptively is by the comprehensive consideration of the image contents and buffer state.This algorithm predicts the bitrate and quality of the image by source video model which is deduced according to the rate-distortion theory and used to describe the relationship of the bitrate, distortion and quantization step, and thus decides the number of skip frames.TMN8 infers the predicted formula of the target bitrate according to the experience of entropy model, then refers to the rate-distortion model, then computes the optimum quantization step under the MSE rule by Lagrange optimization.VM8 is based on quadratic R-Q model, and uses the model in different types of image frames to achieve rate control, meanwhile introduces sliding window to adjust the parameters of the model in order to realized multi-scale, different complexity rate control.In 2001, he (Zhihai He, 2001) proposed a ρ domain code rate control algorithm; it establishes the one-to-one correspondents between output rate and the quantification step by the linear relationship of the percentage of the quantified DCT coefficients and the output rate.This algorithm has achieved good results in the standard of JPEG, H.263, MPEG-4 and so on.The latest video coding H.264 standard in the code control is proposed by Li Zhengguo etc in 2003.The problem with the JM H.264 encoder lies with the fact that the residual signal depends on the choice of coding mode and the choice of coding mode depends on the choice of QP which in turn depends on the residual signal (a chicken and egg type of problem).The adopted solution in the JM encoder is one where the choice of QP is made prior to the coding mode decision using a linear model for predicting the activity of the residual signal of the current basic unit (e.g.frame, slice, macroblock) based on the activity of the residual signal of past (co-located) basic units.Once the residual signal activity is predicted, the same rate model used in VM8 is employed to find a QP which will lead to a bit stream that adheres to the specific bit budget allocation and the buffer restrictions.In order to get a better effect on rate control, we usually make some melioration based on the joint scalable video model (JSVM).The JSVM provides a rate control scheme, and the JSVM software is the reference software for the Scalable Video Coding (SVC) project of the Joint Video Team (JVT) of the ISO/IEC Moving Pictures Experts Group (MPEG) and the ITU-T Video Coding Experts Group (VCEG).The JSVM Software is still under development and changes frequently. The key technique in rate controlBecause of transmission bandwidth and storage space limitation, video applications for higher compression ratio, nondestructive coding can provide the compression ratio but cannot satisfy the demand of actual video applications, but if we can accept some degree of distortion, high compression ratio is easy to get.Human visual system for high frequency signals change not sensitive information loss, high frequency part does not reduce subjective visual quality.Video coding algorithm of mainstream DCT quantization method is adopted www.intechopen.comRate Control in Video Coding 81 to eliminate video signals, the visual physiology redundant than lossless higher compression ratio and will not bring the video quality decrease significantly.When using a lossy coding method, it is related to the difference between the reconstruction images g (x, y) and the original image f (x, y).Generally, the distortion factor D function can form according to need, such as selecting any cost function, absolute square cost function, etc.In the image coding D is computed as: Rate distortion modelBeneath the image compression, there is a problem: under the premise of certain bitrate, how to make the distortion of the reconstructed image coding minimum.Essentially, it is the problem of the relationship between encoding rate and the distortion.The rate-distortion theory is to describe the relations of the distortion of coding and encoding speed.Although the rate-distortion theory is not optimal encoder, but it gives the lower compression allows under the condition of the certain information distortion allows.Practical application of many rate-distortion models is built on the basis of experience.For example, in TM5, a simple linear rate-distortion model is introduced.In TMN8 and VM8, a more accurate quadratic R-D model is used, which can reduce rate control error and provide better performance but have relatively higher computational complexity.In a different way, the relation between rate and QP is indirectly represented with the relation between rate and ρ, where ρ is the percent of zero coefficients after quantization; and also, a modified linear R-D model with an offset indication overhead bits is used for rate control on H.261/3/4 in the contributions.Here are some of the common empirical models: How to referenceIn order to correctly reference this scholarly work, feel free to