Variable and constant bitrate in a DVC to H.264/AVC transcoder

https://doi.org/10.1016/j.image.2011.05.001Get rights and content

Abstract

Mobile-to-mobile video communications constitute one of the main research areas dealing with the dynamic adaptation of traffic generated by video sources. In a framework where one mobile device sends video information to another, both transmitter and receiver should employ video encoders and decoders with low complexity. In this paper, a Variable/Constant Bitrate DVC to H.264/AVC Transcoder is proposed which takes the advantage of both paradigms in terms of low-complexity algorithms on the end-user device side (DVC encoder and H.264/AVC decoder). The proposed transcoder is based on the hypothesis that common DVC GOPs can be converted to H.264/AVC GOPs without significant rate–distortion and bitrate losses, in a flexible way. An in-depth study of the different frame types available in DVC has been carried out in order to exploit the correlation between them and the most suitable GOP pattern in H.264/AVC. Moreover, a dynamic motion estimation technique is proposed in this paper for optimizing the search area for the motion vectors, with the purpose of being used in combination with the GOP mapping approach. Simulation results show that the proposed approaches reduce the DVC to H.264/AVC transcoder complexity by up to 60% on average, while maintaining the coding efficiency in CBR and VBR scenarios, achieving very high quality results over different types of metrics (both objective and subjective). Finally, we conduct a comparative study with all the most prominent DVC transcoding proposals available in the literature, showing that the proposed transcoder achieves the best results (in terms of PSNR and bitrate).

Highlights

► We propose a fast DVC to H.264/AVC transcoder for mobile video communications. ► The transcoding complexity is reduced up to 60% maintaining the quality/bitrate. ► The proposed transcoder permits the conversion between different frame structures.► The proposed transcoder is tested with constant and variable bitrate frameworks.

Introduction

Multimedia communication between mobile devices is becoming an important area of interest in telecommunications due to the advance in mobile networks (such as 4 G). This latest technology gives us better communication capabilities, such as Quality of Service and higher bandwidth savings, among others. Although the network can support the transmission of video content, mobile devices (sender and receiver) require low-complexity encoding and decoding mechanisms. On one hand, standard codecs, such as H.264/AVC [1], provide decoders which are less complex than encoders, since spatial and temporal correlations are exploited on the encoder side. On the other hand, the Distributed Video Codec (DVC) [2] offers a new paradigm where temporal correlation is not exploited on the encoder side, permitting DVC encoders which are less complex than decoders. Consequently, in order to support mobile-to-mobile video communications in an efficient way, we propose the introduction of a transcoder into the network, which converts from DVC to H.264/AVC by taking advantage of both paradigms in order to reduce the complexity in the end-user devices by moving the high-level computational tasks to that transcoder (the DVC decoder and the H.264/AVC encoder). In addition, in a real framework, a good transcoder should have flexibility as well as good performance. However, between DVC and H.264/AVC there are many differences, such as frame type, Group of Pictures (GOP) patterns, and GOP sizes, and these need to be overcome in the transcoder. Consequently, the target in this paper is to provide a DVC to H.264/AVC transcoder which permits transcoding from each DVC GOP to any H.264/AVC GOP efficiently by reusing the Motion Vectors (MVs) generated in the DVC-side information process to accelerate the Motion Estimation (ME) task in H.264/AVC. Furthermore, as is common knowledge, some networks and devices only support a constant data rate. For this reason, as is shown in Fig. 1.1, in this work the proposal has been tested under two different scenarios: Variable Bit Rate (VBR) and Constant Bit Rate (CBR).

Concerning the complexity of the transcoder, this work focuses on reducing the complexity of the DVC to H.264/AVC transcoder (particularly the H.264 stage) by means of reusing DVC information. Although the complexity of the H.264 encoding stage is significantly reduced, the complexity of the first stage (DVC decoding) is still a bottleneck for the whole system. To tackle this issue, our recent work has focused on reducing this complexity through parallel computing [3].

Section snippets

Related work

Many different transcoding approaches based on traditional standards have been proposed in the literature, such as MPEG-2 to H.264/AVC [4], H.263 to H.264/AVC [5], and so on. However, there are only a few approaches based on DVC which have been proposed recently to transcode DVC to H.263 [6] as well as DVC to H.264/AVC [7].

Although the idea of applying the DVC paradigm in a transcoder framework was introduced in [8], it was in 2008 when Peixoto et al. [6] proposed the first architecture to

Flexible DVC to H.264/AVC video transcoder

To provide a framework with low complexity at both ends, it is necessary to convert from a source format with low complexity at the encoder, to another one with low complexity at the decoder. In our architecture (as Fig. 3.1 shows), the first stage is composed of a Wyner-Ziv decoder based on the VISNET-II architecture [10]. Specifically, we have chosen the Transform Domain (TD) because of the better RD results obtained. The Wyner-Ziv codec encodes two types of frames: Key Frames (K) and

Performance evaluation

The proposed DVC to H.264/AVC transcoder has been implemented using a DVC codec based on VISNET-II software [10] and the H.264/AVC JM 14.0 reference software [12]. To test our architecture we have selected the same sequences with the same frame rate that were selected in the DISCOVER codec evaluation [13].

Conclusions and future work

In this work we have presented a low-complexity Variable/Constant Bitrate DVC to H.264/AVC transcoder for mobile-to-mobile video communication. The architecture proposed permits the conversion between common DVC GOPs to the most suitable H.264/AVC GOP, adapting these GOP mappings to the scenario requirements. In addition, a dynamic motion vector optimization algorithm, which reuses the MVs provided by the DVC decoding stage, is also proposed to accelerate the H.264/AVC ME module. As a result,

References (20)

  • C. Brites et al.

    Evaluating a feedback channel based transform domain Wyner-Ziv Video Codec

    Signal Processing: Image Communication

    (April 2008)
  • ISO/IEC International Standard 14496-10:2003, Information Technology – Coding of Audio – Visual Objects – Part 10:...
  • A. Aaron, R. Zhang, B. Girod, Wyner-Ziv coding for Motion Video, in: Asilomar Conference on Signals, Systems and...
  • A. Corrales-García, J.L. Martínez, G. Fernandez-Escribano, Reducing DVC decoder complexity in a multicore system, in:...
  • G. Fernández-Escribano et al.

    A fast MB mode decision algorithm for MPEG-2 to H.264/AVC P-frame transcoding

    IEEE Transactions on Circuits and Systems for Video Technology

    (2008)
  • G. Fernández-Escribano et al.

    Low-complexity heterogeneous video transcoding using data mining

    IEEE Transactions on Multimedia

    (2008)
  • E. Peixoto et al.

    A Wyner-Ziv video transcoder

    IEEE Transactions on Circuits and Systems for Video Technology

    (2010)
  • J.L. Martínez, H. Kalva, G. Fernández-Escribano, W.A.C. Fernando, P. Cuenca, Wyner-Ziv to H.264/AVC video transcoder,...
  • B. Girod et al.

    Distributed video coding

    Proceedings of the IEEE

    (January 2005)
  • T. Wiegand et al.

    Overview of the H.264/AVC Video Coding Standard

    IEEE Transactions on Circuits and Systems for Video Technology

    (2003)
There are more references available in the full text version of this article.

Cited by (5)

This work was supported by the Spanish MEC and MICINN, as well as European Commission FEDER funds, under Grants CSD2006-00046 and TIN2009-14475-C04. It was also partly supported by JCCM funds under Grants PEII09-0037-2328 and PII2I09-0045-9916, and the University of Castilla-La Mancha under Project AT20101802. The work presented was carried out using the VISNET2-WZ-IST software developed in the framework of the VISNET II project. The authors would like to thank Eduardo Peixoto for his valuable support, which helped to improve the manuscript.

View full text