Visual Media Coding and Transmission

3.4 Error Robustness for Scalable Video and Image Coding

3.4.1 Correlated Frames for Error Robustness

3.4.1.1 Problem Definition and Objectives

Scalable video coding supports efficient coding for both intra-frame and inter-frame [82], [83]. A group of coding pictures (GOP) is described, with a hierarchical structure depicted in Figure 3.39. It is noted that the size of GOP is restricted to a dyadic choice [84], i.e. the power of 2, in order to maintain the scalability feature in a temporal fashion. Each GOP contains a key frame, which is defined at the last picture. The coding option for key frames is determined by the frequency of intra-frame, I_Freq. If the value of I_ Freq is equal to that of GOP size, all the key frames are described as I-frames. Otherwise, P-frame coding is employed for non-intrakey frames. Note that the SVC hierarchical structure does not allow more than one I-frame within a GOP. Thus, non-key pictures are encoded as P-frames or B-frames.

The value of a key frame (presence and quality) for decoding the entire GOP structure is significant. A lost key frame, due to poor channel conditions, results in a dysfunctional decoding outcome for GOP reconstruction. This causes quality degradation, noticeable frame-freezing, and disturbing motion-jerking phenomena [85]. Hence the design of error robustness for the key frames is essential in error-prone transmission to ensure a good quality of service (QoS). The error-concealment strategy supported by the SVC utilizes ...