IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY, VOL. 18, NO. 1, JANUARY 2008 59 Redundant Slice Optimal Allocation for H.264 Multiple Description Coding Tammam Tillo, Member, IEEE, Marco Grangetto, Member, IEEE, and Gabriella Olmo, Senior Member, IEEE Abstract—In this paper, a novel H.264 multiple description tech- nique is proposed. The coding approach is based on the redundant slice representation option, defined in the H.264 standard. In pres- ence of losses, the redundant representation can be used to replace missing portions of the compressed bit stream, thus yielding a cer- tain degree of error resilience. This paper addresses the creation of two balanced descriptions based on the concept of redundant slices, while keeping full compatibility with the H.264 standard syntax and decoding behavior in case of single description recep- tion. When two descriptions are available still a standard H.264 decoder can be used, given a simple preprocessing of the received compressed bit streams. An analytical setup is employed in order to optimally select the amount of redundancy to be inserted in each frame, taking into account both the transmission condition and the video decoder error propagation. Experimental results demon- strate that the proposed technique favorably compares with other H.264 multiple description approaches. Index Terms—H.264, multiple description coding, rate alloca- tion. I. INTRODUCTION V IDEO compression is mainly based on block-based motion prediction and compensation, which exploits the temporal correlation between subsequent frames. However, the presence of a prediction loop in the video codec makes the compressed sequence highly vulnerable to errors, due to the dependency among successive coded frames. As a consequence of the pos- sible error propagation, video transmission over networks sub- ject to packet losses is very challenging. In order to overcome this vulnerability, one can consider the use of multiple descrip- tion coding (MDC), where different non hierarchical represen- tations (or descriptions) of the same data, yielding mutually re- finable information, are generated in order to be transmitted over independent paths. In this paper we will limit ourself to the prac- tical situation of two descriptions. The most popular methods to generate MD are based on the pioneering MD scalar quantizer (MDSQ) proposed in [1]; this principle has been applied to video coding in [2]. Another class of methods employs correlating transforms [3]; this approach has been applied to motion compensated MD video coding in [4]. These methods, although providing good performance, are Manuscript received November 11, 2005; revised August 21, 2006 and Feb- ruary 22, 2007. This work has been supported in part by the European Union under NewCom Network of Excellence. This paper was recommended by As- sociate Editor H. Sun. T. Tillo and G. Olmo are with Dipartimento di Elettronica, Politecnico di Torino, 10129 Torino, Italy (e-mail: tammam.tillo@polito.it; marco.grangetto@ polito.it). M. Grangetto is with Dipartimento di Informatica, Universita’ degli Studi di Torino, 10149 Torino, Italy (e-mail: gabriella.olmo@polito.it). Digital Object Identifier 10.1109/TCSVT.2007.913751 conceived as stand alone codecs, and generate descriptions that are incompatible with video standards such as H.264 [5]. This consideration has led to the proposal of MDC schemes that can be configured as pre- and post-processing stages to be associ- ated with any video co-decoder. Among them we can mention [6]–[8], where the correlation among descriptions is obtained by means of 1-D or 2-D oversampling of the original image by zero padding in the DCT transform domain; the inverse trans- form is then operated, and the resulting image is split into subim- ages representing the descriptions. In [9], the 1-D zero-padding scheme proposed in [10] for images is generalized to video, with a frame by frame approach; the scheme is shown to outperform 2-D oversampling for video, as explained in [11]. In [12], the authors suggest to use the slice group coding tool available in H.264 in order to create two balanced description. Although the generated descriptions are indeed H.264 compliant, the use of the slice group modality impairs the compression efficiency. In order to mitigate this effect, in [13] the authors suggest using three-loop slice group MDC; nevertheless this latter solution still exhibits a performance impairment in terms of coding effi- ciency at the two side encoders. In this paper we propose a novel MDC coding approach that generates descriptions that are fully compliant with the H.264 video coding standard. This goal is achieved by exploiting the redundant slice coding option available in the standard. Never- theless, the proposed approach is general and can be used with any other hybrid video codec even if, in such a case, compati- bility with the original standard definition may not be guaran- teed. The most important contributions of the present work are the derivation of an optimal redundancy allocation strategy and its implementation within an open loop rate control technique, based on the selection of the quantization parameter. It is worth noticing that the proposed technique is able to optimally allo- cate the MDC redundancy according to the network status and the error propagation characteristic of the employed codec. The rest of the paper is organized as follows. In Section II, the basic MDC principles are recalled. The proposed method is described in Section III. Section IV presents the analytical optimization approach and its actual implementation using the H.264 syntax. In Section V results are presented. Finally, in Sec- tion VI some conclusions are drawn. II. PRINCIPLES OF MULTIPLE DESCRIPTION In the MDC approach, two or more independently decodable representations (descriptions) of the same data are generated. If these descriptions are transmitted over independent channels and if any subset of them is received, the side decoder can re- construct the data with a given side distortion. The more de- 1051-8215/$25.00 © 2008 IEEE