The influence of interlacing on source coding efficiency and motion estimation

Interlacing is undoubtedly suited for analogue transmission of standard TV since transmission bandwidth is halved while vertical resolution is, to some extent, maintained. On the other hand, however, interlacing has some severe drawbacks as regards the desired convergence of digital video, computers and telecommunications, not to mention the fact that image quality is restricted in principle by the well-known interlace artefacts. Another aspect is that transmission bandwidth in an all-digital system does not primarily depend on the source rate but rather on the transmission rate. Since source statistics of interlaced and progressive scanned image sequences differ significantly, source coding efficiency might be different as well. If interlacing is applied as a means of bandwidth reduction, it has to be considered as an integral part of the compression system. The same applies for the interlace artefacts which are part of the overall “coding error”. In the paper the rate-distortion efficiency of interlaced image coding is analysed for motion-compensated interframe/interfield coding. The theoretical results are verified by using codec-simulations and subjective tests. As an important theoretical and practical result the author finds that motion-compensated prediction with fractional pel accuracy exhibits poor performance as regards prediction gain and motion estimation accuracy when applied to interlaced images. At some transmission rates the author observes strong subjective preference for the progressive scan in the subjective tests when motion-compensated hybrid coding is applied.