Comparisons between the one-loop and two-loop solutions for improving the coding efficiency of FGS

As the MPEG core experiments have shown, the current FGS video coding scheme suffers from a 2~3dB coding efficiency loss compared with the non-scalable single-layer coding for a wide range of bit rates. This is mainly due to the open-loop enhancement structure that ensures drifting free coding. Two categories of schemes are proposed to address this issue with the same basic idea, i.e., introducing some enhanced reference layers into the motion compensation loop. In the one-loop solution, several enhancement bit planes are directly introduced to the motion compensation loop of the base layer. On the other hand, the two-loop solution establishes another motion compensation loop at the enhancement layer. There are many different characteristics with these two solutions. This paper analyzes the one-loop and two-loop solutions in detail from different respects, including coding efficiency, drifting errors, flexibility, and complexity. In the end, the experimental results at the same conditions verify the differences between the oneloop and two-loop solutions.