Font Size: a A A

Research On Key Technologies At The Encoder In Distributed Video Coding

Posted on:2017-11-06Degree:DoctorType:Dissertation
Country:ChinaCandidate:C Y HuFull Text:PDF
GTID:1318330566954650Subject:Information and Communication Engineering
Abstract/Summary:PDF Full Text Request
Video which is the most commonly used and important information carrier in human's social activities is the lively presentation of human's visual perception of the objective things.It's also the main way for us to access to the external information.Along with the development of the communication network and microelectronics technology,some emerging applications such as low power wireless video monitoring network and video sensor network have been rapidly developed.These new applications due to the constrained resources and limited power consumption are unsuitable to adopt the traditional video coding.Distributed video coding(DVC)due to its low-complexity encoder,good rate distortion performance and better error resilience robustness has received many attentions.After decades of research,DVC has made great progress in the areas such as the framework of the system,the video compression efficiency,the rate distortion(RD)performance and so on.But the improvements are at the cost of increasing the system complexity.Especially the techniques at the encoder,such as quantization technique,encoder block mode decision(EBMD),encoder rate control(ERC),encoder side information(ESI)production,will increase the complexity and bring the computation load at the encoder which don't match the low-complexity nature of the encoder and affect the practicality of the DVC system.This paper focuses on the distributed residual video coding(DRVC)system and finds that the encoder can keep low complexity when solve the problems of EBMD and ERC.The main work and the research results obtained include:1.According to the statistical characteristics of the residual signal in DRVC system,the non-uniform quantization scheme of the residual video signal is obtained based on the experimental results.At the same time,the statistical characteristics of the quantized residual signal at the encoder can approximately estimate the correlation between the signals both at the encoder and the decoder.2.A low-complexity and efficient EBMD of the residual frame(EBMD-RF)is proposed.The exiting EBMD algorithms have two disadvantages: one is the block mode decision often depends on the thresholds or the initial parameters,but the user doesn't have a clue how to set these values.The other is the metrics computed for block mode decision will bring the computational burdens at the encoder.The proposed EBMD-RF only depends on the value of the residual pixels at the encoder to decide the macro blocks.EBMD-RF is very simple without any heavy computation and the threshold setting is simple and convenient.These advantages guarantee a low-complexity encoder.The experiment results show EBMD-RF is efficient in mode decision which results in competitive RD performance when compared with DISCOVER system.3.An encoder rate estimation scheme for DRVC(ERC-DRVC)is proposed which consists of two parts,namely a residual ERC algorithm based on pseudo-random code scrambling(RERC-PRCS)and a quantization index estimation algorithm based on pseudo-random code scrambling(QIEA-PRCS).The existing ERC algorithms have two disadvantages: one is the algorithms of generating the accurate ESI are usually complex which will increase the encoder complexity.The other is that the rate is often estimated at the bit plane level which will bring computational load and latency when there are many bit planes.The proposed RERC-PRCS algorithm is used to estimate the rate at the frame level and the QIEA-PRCS algorithm is used to solve the problem of underestimation.ERC-DRVC scheme does not carry on the rate estimation at the bit plane level and only uses the average interpolation algorithm to generate the reference frame,which ensures the low complexity of the encoder.The simulation results show that ERC-DRVC achieves better and competitive performance in the videos with slow motion when compared with common ERC schemes and DISCOVER system,respectively.While in the videos with high motion,ERC-DRVC achieves the competitive performance when compared with the common ERE schemes and has a RD gap when compared with DISCOVER system.4.A more effective and practical ERC-EBMD-DRVC is proposed.At the encoder,all blocks in the residual frame are classified into Skip blocks,Intra blocks and WZ blocks by using EBMD-RF algorithm.Then the transmitting rate for all the WZ blocks is estimated by using RERC-PRCS algorithm and the problem of the rate underestimation is also solved by QIEA-PRCS algorithm.The simulation results show the proposed ERC-EBMD-DRVC outperforms DISCOVER and the state-of-the-art ERC solution in the videos with good side information and has a little worse RD performance than the two schemes in the videos with unsatisfactory side information.Due to the simple framework and the low-complexity encoder,our solution is promising in practice.
Keywords/Search Tags:distributed residual video coding, encoder block mode decision, encoder rate control, scrambling with pseudo-random codes, low-complexity encoder
PDF Full Text Request
Related items