Pamela Zontone

Rate allocation for robust video streaming based on distributed video coding

This paper proposes an error resilient coding scheme that employs distributed video coding tools. A bitstream, produced by any standard motion-compensated predictive codec (MPEG-x, H.26x), is sent over an error-prone channel. In addition, a Wyner-Ziv encoded auxiliary bitstream is sent as redundant information to serve as a forward error correction code. At the decoder side, error concealed reconstructed frames are used as side information by the Wyner-Ziv decoder, and the corrected frame is used as a reference by future frames, thus reducing drift. We explicitly target the problem of rate allocation at the encoder side, by estimating the channel induced distortion in the transform domain. Rate adaptivity is achieved at the frame, subband and bitplane granularity. Experimental results conducted over a simulated error-prone channel reveal that the proposed scheme has comparable or better performance than a scheme where forward error correction codes are used. Moreover the proposed solution shows good performance when compared to a scheme that uses the intra-macroblock refresh procedure.

Frame-based multiple-description video coding with extended orthogonal filter banks

We propose a frame-based multiple-description video coder. The analysis filter bank is the extension of an orthogonal filter bank which computes the spatial polyphase components of the original video frames. The output of the filter bank is a set of video sequences which can be compressed with a standard coder. The filter bank design is carried out by taking into account two important requirements for video coding, namely, the fact that the dual synthesis filter bank is FIR, and that loss recovery does not enhance the quantization error. We give explicit results about the required properties of the redundant channel filter and the reconstruction error bounds in case of packet errors. We show that the proposed scheme has good error robustness to losses and good performance, both in terms of objective and visual quality, when compared to single description and other multiple description video coders based on spatial subsampling. PSNR gains of 5 dB or more are typical for packet loss probability as low as 5%.

Cross-Layer Joint Optimization of FEC Channel Codes and Multiple Description Coding for Video Delivery over IEEE 802.11e Links

The paper presents two cross-layer optimization strategies based on the IEEE 802.11e standard that enable a robust video transmission using adaptively forward error correction (FEC) channel codes at transport layer and a multiple description coding (MDC) architecture. The first approach estimates an array of correlation measures for the group of picture to be transmitted and classifies it using a vector quantizer. The most appropriate coding mode is selected according to the correlation class and the channel conditions obtained through a cross-layer signalling protocol. The second solution relies on a parametric model of the channel distortion based on the percentage of null quantized transform coefficients, which can be obtained during the coding operations. According to the state of the network, the optimization algorithm estimates the expected quality of the reconstructed sequence for each mode and chooses the best one. Experimental results show that both algorithms perform well with a small computational effort but different playout delays.

Cross-Layer joint optimisation of FEC channel codes and Multiple Description Coding for video delivery over IEEE 802.11e links

The paper presents two cross-layer optimization strategies based on the IEEE 802.11e standard that enable a robust video transmission using adaptively forward error correction (FEC) channel codes at transport layer and a multiple description coding (MDC) architecture. The first approach estimates an array of correlation measures for the group of picture to be transmitted and classifies it using a vector quantizer. The most appropriate coding mode is selected according to the correlation class and the channel conditions obtained through a cross-layer signalling protocol. The second solution relies on a parametric model of the channel distortion based on the percentage of null quantized transform coefficients, which can be obtained during the coding operations. According to the state of the network, the optimization algorithm estimates the expected quality of the reconstructed sequence for each mode and chooses the best one. Experimental results show that both algorithms perform well with a small computational effort but different playout delays.

Performance evaluation of wavelet-based distributed video coding schemes

In this paper we propose and compare different distributed video coding (DVC) schemes based on the use of the wavelet transform, which naturally allows for spatial and other forms of scalability. In particular, we propose a hybrid encoder which utilizes channel codes, and evaluate its performance in the absence of a feedback channel. The proposed scheme uses statistical models for the estimation of the required bitrate at the encoder. We also propose a scheme that is based on a modulo reduction procedure and does not use channel codes at the receiver/transmitter. These schemes are compared with more conventional coders that do not or only partially exploit the distributed coding paradigm. Experimental results show that the considered schemes have good performance when compared with similar asymmetric video compression schemes, and that DVC can be an interesting option in appropriate scenarios.

Performance evaluation of distributed video coding schemes

Distributed video coding (DVC) has recently been proposed for emerging scenarios, where sources of correlated video do not communicate, as in some video-surveillance applications, or to simplify the coder for video equipment with power consumption constraints. In this paper we propose and compare different DVC schemes. In particular, we propose the use of the wavelet transform and analyze the performance of encoders based on turbo-codes and on a novel modulo-reduction procedure. The proposed schemes do not need feedback from the receiver and use statistical models for the estimation of the required bit-rate. Experimental results show that the proposed schemes have good performance when compared with similar asymmetric video compression schemes, and that DVC can be an interesting option in appropriate scenarios.

Multiple description for robust Scalable Video Coding

Scalable Video Coding (SVC) was recently standardized as an extension of the H.264/AVC standard. A scalable coder allows to combine different layers of spatial, temporal and quality scalability, and is a viable solution for adaptation to user characteristics and network conditions. On the other hand, Multiple description coding (MDC) can provide error resilience and graceful quality degradation for transmission over error-prone channels. In this paper, we propose a video codec that combines the SVC and MDC coding paradigms. In particular, the enhancement data of the SVC coder are transmitted using MDC. The resulting fully compatible coder takes advantage of the efficiency of SVC and of the robustness of MDC techniques. The effectiveness of the proposed solution is confirmed by experiments.

Bit allocation and quantizer optimization in multiple description coding with oversampled filterbanks

Multiple description coding has been recently proposed for images and video transmission in packet erasure channels because of its good loss recovery capability. In this paper, we consider the problem of bit allocation and quantization optimization for frame based multiple description coding schemes obtained by means of an oversampled filterbank. We generalize the well known solution of orthogonal filterbanks, where the same quantizer should be used in each channel, to the case of oversampled filterbanks with coefficient losses. The resulting design procedure can be used to adaptively adjust the quantizers to the channel conditions. We show that the proposed approach is beneficial by means of experiments with images and video.

Distributed Video Coding: Principles and Evaluation of Wavelet-Based Schemes

The current scenario considered for video coding and transmission, as assumed by the MPEG standards, uses complex encoding algorithms, including motion compensation and multi-hypothesis rate-distortion optimized procedures, to achieve high compression efficiency. This is the right solution when the encoding is carried out by high performance devices, while the receivers (e.g., handheld devices or mobile phones) should be kept as simple and cheap as possible. This scenario is rapidly evolving into a new one, where users have the interest to produce and transmit video and multimedia, possibly using their mobile and battery operated lightweight devices. Of course, this calls for new multimedia coding paradigms, where the encoder is as simple as possible to reduce the computational power requested for compression and radio transmission. Similar constraints should be considered in monitoring or surveillance applications, where a number of video sensors, with limited computational power, cooperate to send video information to a receiving station. Distributed Source Coding (DSC) refers to the compression of two or more correlated sources that do not communicate with each other (hence the term distributed coding). These sources send their information to a central decoder that performs joint decoding. In this situation, the challenging problem is to achieve the same efficiency (the joint entropy of correlated sources) while not requiring sources to communicate with each other. The Slepian-Wolf Theorem is a celebrated result of information theory which assures that this unexpected result is indeed achievable. In other words, it is possible to code two correlated sources (X,Y) one independently of the other, and achieve the same performance obtainable when a coder can exploit knowledge of both. For instance, in a conventional video coder, two consecutive frames can be compressed by computing the motion compensated difference between the first and the second frame, then transmitting the first frame in intra-mode and the inter-mode coded difference frame. The Slepian-Wolf result assures that, in principle, we can achieve the same coding efficiency by coding the two frames independently, i.e., without performing a costly motion compensation operation.