Accounting for quantization noise in online correlation noise estimation for Distributed Video Coding

Abstract

In Distributed Video Coding (DVC), compression is achieved by exploiting correlation between frames at the decoder, instead of at the encoder. More specifically, the decoder uses already decoded frames to generate side information Y for each Wyner-Ziv frame X, and corrects errors in Y using error correcting bits received from the encoder. For efficient use of these bits, the decoder needs information about the correlation between X available at the encoder and Y at the decoder. While several techniques for online estimation of correlation noise X − Y have been proposed, the quantization noise in Y has not been taken into account.As a solution, in this paper, we calculate the quantization noise of intra frames at the encoder and use this information at the decoder to improve the accuracy of the correlation noise estimation. Results indicate average Wyner-Ziv bit rate reductions up to 19.5% (Bjøntegaard delta) for coarse quantization.

Compensating for Motion Estimation Inaccuracies in DVC

“…A GOP of size 8 is used, and only the luma component is coded for comparison with the state-of-the-art DISCOVER codec 4 . Our system is also compared to our previous work [7], which applies a better correlation noise model than DISCOVER, but still uses only one predictor per block. The results indicate improvements over both systems (Fig.…”

“…The proposed codec is based on the work of Aaron et al [1], with some important extensions adopted from DISCOVER [2], and from our previous work [7]. The codec is depicted in Fig.…”

“…to calculate the quantization noise, which is needed at the decoder for accurate correlation noise estimation, as in [7]. WZ frames are partitioned into 4-by-4 nonoverlapping blocks, and each block is transformed using a DCT.…”

“…The scaling parameter α is calculated based on the reference residual of the linear predictor, using the techniques proposed in our earlier work [7].…”

Flexible distribution of complexity by hybrid predictive-distributed video coding

“…Some important contributions have been made in the context of the DISCOVER project, such as improvements to the generation of the side information by using bidirectional motion refinement and spatial smoothing, and adaptive GOP size control [15] or improvements have been proposed for side information generation, such as exploiting both temporal and spatial correlation [16,17], and using multiple side information streams [18]. Techniques for online modeling of the correlation noise between the original and the side information have been developed [19,20], taking into account the quantization noise in the reference frames [21]. Rate-distortion analysis has been provided for motion extrapolation [22], motion compensated interpolation [23] and hash-based side information generation [24].…”