Efficient weighted prediction parameter signaling using parameter prediction and direct derivation

Abstract

Weighted prediction (WP) is an efficient motion compensation technique to predict temporal illumination variation. It has been introduced in H.264/MPEG-4 AVC (Advanced Video Coding, H.264) and consists of two modes: an explicit WP (EWP) and an implicit WP (IWP). Since in EWP, a weighting factor and an offset for each reference picture are explicitly transmitted to the decoder, an overhead of these parameters affects the coding efficiency. Since in H.264-based EWP, these parameters are directly encoded without prediction, it is difficult to apply WP for low bitrate applications. In order to reduce the overhead of them, this paper presents an efficient weighted prediction parameter signaling method towards High Efficiency Video Coding (HEVC) standard by introducing a WP parameter prediction and a direct WP parameter derivation. A weighting factor and an offset are predicted using the characteristics of WP parameters and entropy-coded. Moreover, when there are the same entries in reference pictures for prediction list 0 and 1 in the case of B-slices, a set of WP parameters for prediction list 1 is directly derived from that for prediction list 0. The experimental results show that our method can achieve up to 66% overhead bitrate reduction and up to 7% overall bitrate reduction compared to the conventional H.264-based EWP.

Efficient temporal and interlayer parameter prediction for weighted prediction in scalable high efficiency video coding

“…In Ref. 27, WP parameter prediction is further enhanced for each reference frame in list 1 by deriving the weight and offset from those reference frames in list 0 . For each reference frame in list 1 , when the estimated WP parameters, W COMP L S ;list 1 ;j and O COMP L S ;list 1 ;j , are exactly the same as those of one reference frame in list 0 , then, the WP parameters can be directly reused by signaling that the particular reference index in list 0 .…”

“…Prediction To enhance IWPP 26 and enhanced direct WP parameter derivation (DWPD), 27 lossless and more efficient WP parameter prediction in both of the base layer, L B , and the enhancement layer, L E , is proposed, which has interlayer WP parameter prediction in L E as well.…”

“…As aforementioned in Ref. 27, the reference frame with the same weight and offset of the reference frame in list 0 needs the additional signaling bits to indicate which reference frame in list 0 would be employed for reusing of WP parameters. As an alternative, we would retrieve the WP parameters by checking the POC of each reference frame in list 0 so that additional signaling bits are not required.…”

“…The coding of W COMP L S ;list 1 ;j and O COMP L S ;list 1 ;j can be skipped. Compared with DWPD, 27 the additional signaling bit and the reference index difference in Eq. ( 8) can also be saved, which can improve the coding efficiency.…”

“…Parameter Derivation Versus Enhanced Direct WP Parameter Derivation in Base Layer For DWPD, 27 it can achieve 0.33% and 1.53% average bitrate reduction for RA and LD, respectively, compared with HWPP and with corresponding 15.27% and 41.07% overhead reduction. It is noted that there is no bitrate reduction or overhead reduction for DWPD since there is no list 1 reference for the LDP configuration.…”