The broadcast distribution network is becoming ever more complicated due to the ongoing proliferation of acquisition, processing and distribution standards. Departing from the model of simple network-to-home TV transmissions, today's media marketplace allows consumers to choose from an immense range of delivery mechanisms, display equipment and support media. In providing this wealth of options, media companies face added challenges in maintaining the quality of their assets and services.
Every link in the distribution chain carries the risk of compromising the quality of content. Poor quality not only reduces the value of content, but also undermines satisfaction. However, with careful attention to video processing at every point in the distribution chain, content owners, aggregators, service providers and infrastructure managers can preserve the best possible image quality and, in turn, maintain their own efficiency and profitability.
Keying in on conversion
Signal distribution in the broadcast environment typically includes the ingest of media, editing or repurposing of that content, as well as archiving and/or delivery over-the-air or via cable, satellite, IPTV service, Web portal, or mobile network. In any version of this distribution chain, the broadcaster's ability to meet high-level requirements depends on accurate and efficient signal management and processing. Synchronization, maintenance of signal levels, contribution and storage compression, clean switching, and transparent conversion all contribute to the value of the end product. Among these, conversion is particularly important because it is present — visible and invisible — in so many areas of the signal distribution chain.
The multitude of high-resolution, low-resolution, progressive scan and interlace scan picture formats used across the global broadcast market are reconciled through a variety of conversion processes. Across the distribution chain, deinterlacing, format conversion, standards conversion and aspect ratio conversion occur at many easily identifiable points.
Most conversion processes involve manipulation of interlaced material. Even a simple aspect ratio conversion is best performed in the progressive domain, after a high-quality deinterlace.
Interlaced video provides a compromise between data rate, temporal resolution and vertical resolution, and this compromise can be seen as a primitive form of compression. The upside is the potential for higher spatial resolution of still images. The downside is that as soon as anything in the picture moves, there is ambiguity between vertical resolution and temporal motion. As a result, the quality of any conversion process — and a number of other image restoration and enhancement operations — depends on an understanding of the true motion in the picture, and this understanding is achieved through motion compensation.
Addressing the issue of motion
Motion compensation is best thought of as being a method by which the relevant filtering can be skewed to match the motion of the picture. The application of motion compensation to standards and format conversion can make a much better job of some of the processing that traditionally has been the preserve of linear or motion-adaptive filtering. An illustration of vertical tilt frequency response, at rest and at critical speeds, demonstrates why linear filters cannot distinguish between vertical detail and motion. In essence, it's because the spectrum skews.
Figure 1 on page 30 shows the vertical/temporal energy distribution for a low-motion, vertically detailed 1080p video sequence. In this case, a simple 2D vertical temporal linear filter (grey box) can easily isolate the green “real” image for further image processing.
Figure 2 shows the same graph for a high-vertical-motion, vertically detailed 1080p video sequence. The simple 2D vertical temporal linear filter (grey box) can no longer easily isolate the green “real” image for further image processing without losing some of the green spectra or including some of the red. Motion compensation allows a skewed filter aperture to be fit to the skewed energy spectra, in turn enabling effective isolation of the “real” image again.
Effective motion compensation, without unsightly artifacts, hinges on the accuracy of the motion vectors generated for an output picture. Phase correlation is a motion-compensation technique that measures “true motion” in the scene rather than aiming merely to identify matching blocks between frames.
Measuring motion accurately
Phase correlation improves on block-matching techniques for motion compensation by actually measuring the speed and direction of moving objects rather than estimating, extrapolating or searching for them. While block-matching techniques excel in identifying areas of similarity, they fail through their assumption that these areas of similarity are the same area, but moved between frames. Thus, used purely for compression, in which motion vectors don't need to be correct, block matching is an effective tool. However, with deinterlacing or standards conversion, these vectors do need to be accurate, as they are used to move pixels to make entirely “new” pictures.
In taking on challenging material, such as repetitive structures or small, fast-moving objects like scrolling credits or a putt on a golf green, phase correlation succeeds because it can match pixels in each field precisely with their counterparts in adjacent fields. The resulting images yield the smooth motion portrayal that is essential to high-quality conversion.
Phase correlation performs a spectral analysis on the two successive fields (frames) by using a Fast Fourier Transform (FFT) and passes the output through an inverse FFT to generate a correlation surface. This surface uses the difference in phase information from field to field to reveal the pixels whose positions correspond to detected motion between the fields. Thus, the distance and direction of the motion are measured accurately, and application of a subsequent object matching process creates the actual vector field for each pixel.
Enhancing service quality and efficiency
While the subjective quality of converted images is key to ensuring customer satisfaction, the technical quality of conversion has a direct impact on the quality of compression and, in turn, the efficiency of the distribution chain. Good compression enables cost-effective transport and storage of high-resolution material, reducing the amount and cost of bandwidth required for signal distribution.
Material flawed with conversion artifacts puts a greater burden on compression than does cleanly converted material. Because bits are unnecessarily dedicated to these artifacts, the compressed material requires more bandwidth and/or suffers from degradation in picture quality. As further processing takes place along the distribution chain, these issues are amplified.
Figure 3 demonstrates the difference that high-quality motion-compensated conversion can make in the efficiency of a downstream encoding system. A sports clip full of action was standards-converted both with a phase correlation motion-compensated converter and an alternative converter. Long-GOP MPEG-2 encoding was performed on each.
As the graph illustrates, the clip converted using phase correlation motion compensation yielded improved downstream compression performance of significant proportions. Such bandwidth efficiencies can enable greater cost savings or make room for the provision of additional revenue-generating services.
The multiple conversions inherent in the links of processes from content production to content consumption can lead to degradation of quality. This results in dissatisfaction not only on the part of the consumer, but also the service provider carrying the programming and the advertiser, who expects a better-looking vehicle for messaging. Through the technologies of phase correlation motion compensation, the best possible quality can be preserved within standards and format conversions.
Gerard Phillips is general manager — conversion with Snell.