|
Framework of error-resilient video coding using flexible macroblock ordering and error concealment for wireless video transmission |
|---|---|
| รหัสดีโอไอ | |
| Title | Framework of error-resilient video coding using flexible macroblock ordering and error concealment for wireless video transmission |
| Creator | Jantana Panyavaraporn |
| Contributor | Supavadee Aramvith |
| Publisher | Chulalongkorn University |
| Publication Year | 2552 |
| Keyword | Video compression, Digital video, Coding theory, Image transmission |
| Abstract | The H.264/AVC standard adopts a robust error resilience tool at the encoder known as Flexible Macroblock Ordering (FMO). The main goal of this tool is to provide a macroblock-level interleaving tool to spread out consecutive burst errors in a frame. Past research proposed the use of macroblock-coded bit-count which acts as spatial information as indicator of macroblock importance, and uses a two-pass encoding process to generate the macroblock-address map. In this dissertation, we propose to use a distortion measure based on concealment error which acts as temporal information as an indicator for a choice of macroblock-address-map of each picture. To avoid the incurred delay and complexity computing of two-pass encoding, we also propose a one-pass encoding scheme to generate the macroblock-address maps. Furthermore, we present a framework that combines one-pass FMO map generation and error concealment algorithms to improve the video quality due to transmission errors. The one-pass FMO map generation is accomplished by using feedback in terms of spatial and temporal information to simulate spatial and temporal error concealment at the encoder. The choice of error concealment method at decoder is applied according to parameter derived from residual information obtained at the encoder during the map generation process. Our simulation results performed under slow and fast fading channel confirm that the proposed technique can reduce the number of undecodable macroblock up to 80.54% and PSNR improvement are up to 6.09 dB when compared with methods that don’t use FMO and uses a simple non-motion compensated error concealment. |
| URL Website | cuir.car.chula.ac.th |
