Joćo Miguel Duarte Ascenso

Fine Grain Scalability

The major problem with scalable video codecs is that they offer good flexibility in terms of scalability, but at the cost of lower efficiency when compared to non-scalable coding. MPEG-4 FGS achieves fine grain adaptation to dynamic changes in network conditions, low complexity decoding and packet loss resilience, but its coding efficiency performance is lower when compared to non-scalable video coding schemes.

This work has been included in my M.Sc. Thesis available here for download (portuguese only).

In this context it has been proposed and published two distinct solutions:

H.264/AVC Fine Grain Scalability using Bitplane Coding (AVC-FGS):

This solution is based on the MPEG-4 FGS bitplane coding scheme, and some of the tools specified in the recent H.264/AVC standard. The AVC-FGS architecture proposed here maintains all the FGS capabilities intact, notably fine grain adaptation to dynamic changes in network conditions, low complexity decoding and memory requirements, and packet loss resilience. In the AVC-FGS enhancement layer, the tools in the H.264/AVC base layer are reused as much as possible to take maximum advantage of them.

J. Ascenso, F. Pereira, “H.264/AVC Fine Grain Scalability using Bitplane Coding”, 5th International Workshop on Image Analysis for Multimedia Interactive Services, Lisbon, April 2004 (paper) (poster) (source code).

H.264/AVC Fine Grain Scalability with Motion Compensated Architecture (AVC-MCFGS):

A fine grain scalability solution has been developed with a two-loop architecture which uses in the base layer a non scalable H.264/AVC encoder. The enhancement layer reuses many base layer tools and includes motion compensation to improve the coding efficiency. Several drift reduction techniques are also proposed to limit the error propagation (drift) when the encoder and decoder references are not the same, due to changes in the available bandwidth or transmission errors. This allows reducing the gap in coding efficiency when compared to the non-scalable state of the art H.264/AVC encoder.

J. Ascenso, F. Pereira, “Drift Reduction for a H.264/AVC Fine Grain Scalability with Motion Compensation Architecture”, IEEE International Conference on Image Processing, Singapore, October 2004. (paper)  (poster)

Research - Teaching - Publications - Links - Contact - Home