Assessment of Flexible Slice Structure

To validate the flexible slice structure, we use Foreman sequence in CIF resolution as an ex-ample. Particularly, according to the shape information, we partition each enhancement-layer frame into the foreground and background slices, as shown in the part (b) of Figure 4.17. The partition information is coded at macroblock level and embedded in the bit-stream. In addition, to make the decoder more robust to errors, the decoding of a slice will stop once an error is detected from illegal semantics. For example, if more than 16 coefficients are found in a 4x4 block, we will discard the remaining bit-stream in the current slice and continue on the decoding of next valid slice.

For generating error pattern, we follow the configuration in Figure 4.19 to periodically insert an error byte (of value 0x00) for every 21 frames. For a fair comparison, the same error pattern is used for different slice structures.

For assessment, Figure 4.20 shows the PSNR comparison on a frame-by-frame basis and Figure 4.21 gives the comparison of subjective quality. As shown in Figure 4.20, the flexi-ble slice structure averagely offers higher PSNR because the error propagation is effectively

Frame

0 21 42 63 84 105 126 147 168 189

PSNR-Y (dB)

15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63

No Error

Fixed Partition Structure with Error

Flexible Slice Structure with Error in Background

Figure 4.20: Luminance PSNR comparison with periodic byte error.

(a) (b)

Figure 4.21: Subjective quality comparison of the 62th frame in Foreman CIF sequence. (a) Flexible slice structure. (b) Fixed partition structure.

Chapter 4. Context-Adaptive Bit-Plane Coding

Frame

0 21 42 63 84 105 126 147 168 189

PSNR-Y (dB)

15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63

No Error

Fixed Partition Structure with Error

Flexible Slice Structure with Error in Foreground

Figure 4.22: Luminance PSNR comparison with periodic byte error.

(a) (b)

Figure 4.23: Subjective quality comparison of the 62th frame in Foreman CIF sequence. (a) Flexible slice structure. (b) Fixed partition structure.

constrained within a slice. Specifically, as shown in the part (a) of Figure 4.21, the errors are constrained in the background slice. On the other hand, with the fixed partition structure, the part (b) of Figure 4.21 illustrates that the errors will propagate to both the foreground and the background regions. In Figures 4.22∼4.23, we show another comparison, in which the fore-ground slice is transmitted prior to the backfore-ground slice and the errors occur in the forefore-ground region.

For the assessment of penalty on coding efficiency, we measure the increase of file size. In this example, the flexible slice structure has only about 0.5% increase on file size, as compared to the fixed partition structure. In fact, the penalty varies with the input sequence and slice structure. More penalties are expected when more slices are created.

4.9 Summary

In this chapter, we propose a context-adaptive bit-plane coding (CABIC) with a stochastic bit reshuffling (SBR) scheme to deliver higher coding efficiency and better subjective quality for fine granular scalable (FGS) video coding. For higher coding efficiency, we construct con-text models based on both the energy distribution in a block and the spatial correlations in the adjacent blocks. Moreover, we exploit the context across bit-planes to save side information.

Experimental results show that our CABIC can provide a PSNR improvement of 0.5∼1.0dB over the VLC based bit-plane coding [15].

Further, for better subjective quality, the proposed SBR incorporates the context probability model to reorder the coefficient bits in such a way that the estimated rate-distortion performance is in descending order. As compared to the approaches with frame raster and coefficient zigzag scanning, our SBR offers better visual quality and also maintains similar or even higher coding efficiency.

In addition, for error resilience, the flexible slice structure partitions the enhancement-layer frames into multiple slices to prevent error propagation. The proposed scheme provides error resilience with little changes on coding efficiency.

This work proves that the bit-plane coding in MPEG-4 FGS [15] can be further improved in coding efficiency and subjective quality. In addition, the flexible slice structure can effectively provide error resilience in the framework of CABIC and SBR. The following summarizes our main contributions in this work.

Chapter 4. Context-Adaptive Bit-Plane Coding

1. We construct context models based on both the energy distribution in a block and the spatial correlations in the adjacent blocks. Moreover, we propose a bit-plane partition scheme to save side information.

2. We use estimated Laplacian distributions to model the transform coefficients and exploit maximum likelihood estimators to minimize the overhead for the Laplacian parameters.

3. While maintaining similar or even higher coding efficiency, our SBR dynamically re-orders the coefficient bits so that the regions containing more energy are coded with higher priority.

4. For the implementation of SBR, we develop a dynamic priority management based on a dynamic memory organization.

5. In addition, we propose a flexible slice structure to provide error resilience in the frame-work of CABIC and SBR.

Applications in Scalable Video Coding Standard

5.1 Introduction

To support clients with diverse capabilities in complexity, bandwidth, power and display reso-lution, the MPEG committee is defining a novel scalable video coding (SVC) framework [23]

that can simultaneously support spatial, temporal, and SNR scalability under the constraints of low complexity and low delay. Currently, the algorithm of SVC [23] is a scalable extension of the newly adopted H.264/AVC video standard [31]. In this chapter, we give an introduction to the SVC algorithm [23] and show how the proposed EMFGS and SBR can be applied in the SVC framework [23] for the improvement of coding efficiency and the functionality of ”region-of-interest”. The rest of this chapter is organized as follows: Section 5.2 presents the SVC framework. Section 5.3 illustrates the application of proposed EMFGS in SVC standard and Section 5.4 demonstrates the applications for SBR. Lastly, Section 5.5 summarizes this chapter.

Chapter 5. Applications in Scalable Video Coding Standard

Decoded Frames {L⁰} Motion

Decoded Frames {L¹} Motion

Core Encoder

Figure 5.1: Encoder block diagram of the scalable video coding standard. [25]