Multiterminal Video Coding

Abstract

Abstract-Following recent works on the rate region of the quadratic Gaussian two-terminal source coding problem and limit-approaching code designs, this paper examines multiterminal source coding of two correlated video sequences to save the sum rate over independent coding. Specifically, the first video sequence is coded by H.264 and used at the joint decoder to facilitate Wyner-Ziv coding of the second video sequence. An efficient stereo matching algorithm based on loopy belief propagation is then adopted at the decoder to produce pixel-level disparity maps between the corresponding frames of the two decoded video sequences on the fly. Based on the disparity maps, side information for both motion vectors and motion-compensated residual frames of the second sequence are generated at the decoder before Wyner-Ziv encoding. Preliminary results on stereo video sequences using H.264 in conjunction with LDPC codes for Slepian-Wolf coding of the motion vectors show savings in terms of the sum rate when compared to separate coding at the same video quality.

Symmetric distributed multiview video coding

“…This is because there is not much correlation after residual coefficients are quantized using a high quantization parameter. The asymmetric distributed multiview video coding scheme in [5] produces the similar results. In the highrate case (QP l = 4, 10), SDVMC-SNS can save 1.62% and 0.59% of total bit rate, respectively.…”

“…In theory, DMVC holds the promise to achieve the same compression performance as JMVC while demanding no communication between cameras. Several distributed multiview video codecs [4], [5] are recently proposed. However, those approaches are based on asymmetric Slepian-Wolf codes which can not achieve the whole Slepian-Wolf rate region and thus limit the rate allocation options between the encoders.…”

Geometry-Based Distributed Scene Representation With Omnidirectional Vision Sensors

“…However, it does not exploit the correlation among I-frames, and, thus, the achieved bit rates are still quite far from the Slepian-Wolf bound. This gap can be reduced by encoding more coarsely the I-frames [17], but a lot of geometric correlation between I-frames is still left unexploited. The work presented in this paper is substantially different from [15] and [17] since it uses epipolar geometry information for the design of the Slepian-Wolf code and, therefore, exploits the correlation between multiview images in a completely distributed manner.…”

“…They take different approaches for modeling the correlation among views, like the disparity-based model [12], affine model [14], or homography-based model [13]. Another direction for the distributed multiview video compression is based on classical motion compensated video encoding at each camera, while the interview correlation is exploited in a distributed manner [15]- [17]. Authors in [15] present a transform-based DSC method for multiview video coding that tracks epipolar correspondences between macroblocks in different views.…”

“…Moreover, the Wyner-Ziv encoder has partial access to the side information (Intra macroblocks and motion vectors), so that this scheme cannot be classified as fully distributed multiview coding scheme. On the other side, a completely distributed stereo-view video coding method is proposed in [16]. It performs independent coding of I-frames and Wyner-Ziv coding of P frames, where the side information is generated by fusing the disparity map with the motion field.…”