Subjective Assessment of a Model-Based Video Codec Compared to H.263

Obtaining a reliable evaluation of the performance of any video codec, especially one designed for very low bit-rate applications, is a difficult task due to the various factors playing a role in the encoding—decoding process — for example, the available channel bit-rate, the frame rate at which the system is required to operate, and most importantly the nature of the encoded scene (scene activity). Each factor introduces radically different artefacts in the pictures, which are difficult to compare and may produce different responses from different people. Nevertheless, there is still a need to assess the performance of such codecs in accordance to the human eye,the very end of any visual communication system. This chapter describes a subjective assessment experiment conducted for evaluating a prototyped model-based coding (MBC) system. However, since there is no other ready-to-use MBC codec that can be comparatively evaluated with the developed one, an H.263 codec is used instead to code the same sequences as coded by the model-based codec. Statistical results obtained from the subjective test show that the MBC codec produced good performance in comparison with the H.263 especially in the very low bit-rate regions.