Font Size: a A A

The Distortion Model Of Reconstructed Picture Considering Reference Frame In Video Coding

Posted on:2013-06-29Degree:MasterType:Thesis
Country:ChinaCandidate:J ChenFull Text:PDF
GTID:2248330371470463Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
Video coding is an important research topic in digital information filed. Various control and optimization techniques based on the rate-distortion models have been adopted in the encoder to improve its performance and satisfy application requirements. Thus a qualified rate-distortion model consisting of the relationship between the bit-rate and the coding parameters and the relationship between the distortion and coding parameters is dispensable to encoders. This thesis establishes a distortion model, a part of rate-distortion model, of reconstructed picture.Reference frame compression (RFC) is a newly research topic in video coding which aims to reduce the memory access bandwidth in high definition real-time video codec. The reconstructed picture is recompressed before it being stored in off-chip memory. However, lossy RFC method will impact the RD performance of video codec. Thus, a distortion model for RFC established in thesis is dispensable as well.Based on the summary of the current rate-distortion models and algorithms of RFC, a distortion model considering the reference frame is proposed. As the current rate-distortion models without taking the RFC into consideration, distortion model for RFC is proposed in this thesis as well.Based on the prediction residuals this thesis establishes the relationship between the distortion and the quantization parameters both of the current frame and the reference frame according to mathematical derivation and data fitting. To solve the RFC problem, the thesis regards the total noise as the sum of two sub coding noises. And the total distortion is derived based on the studies of these two sub-noises and the correlation between them. The distortion model without RFC is modified and improved to fit for the distortion of the first noise and the relationship between energy, mean value and zero probability is concluded for the distortion of the second noise.Finally, the thesis summarizes the methods and the ideas of this research and proposes some research directions worth being further studied.
Keywords/Search Tags:Video Coding, Rate Distortion, Distortion Model, Reference Frame Compression, Quantization Parameter, Coding Noise, Prediction Residual
PDF Full Text Request
Related items