Skip to Main Content
Among many technologies used in wavelet-based scalable video coders, motion compensated temporal filtering (MCTF) is a key technology for both reducing temporal redundancy and providing flexible temporally scalable architecture. However, since its non-flexible temporal pyramid structure introduces a large end-to-end delay of a size of GOP, it is very difficult to use scalable coding technology for many applications that require low end-to-end delay. In this paper, we propose a low-delay unconstrained motion compensated temporal filtering (LD-UMCTF) scheme. Our method focuses to control the end-to-end delay in a unified framework as well as have most useful features of predictive coders, such as multiple reference frame and bi-directional prediction. Experimental results show that the proposed method can provide comparable coding efficiency and has significantly smaller performance degradation compared to the conventional MCTF.