Since the early days of transcontinental and intercontinental networking of TV content, several intermediate communication links have generally been required to complete end-to-end delivery. This is referred to as “concatenation” and has been an area of significant attention regarding the maintenance of video and audio quality and integrity. In earlier analog systems, demodulation and remodulation of the TV signal at each en route microwave or satellite terminal contributed to signal degradation (particularly video signal-to-noise), which, in many instances, reduced the image at the point of delivery to less than that normally acceptable for broadcast.
The transition to video compression and digital networking has created new challenges in mitigating the effects of degradation caused by concatenation in the multiple encode-decode process associated with digital turnaround over satellite, wireless and terrestrial links. One particular problem is maintaining integrity of the 4:2:0 chroma component where an HD-SDI interconnect is used between concatenated decoders and encoders. This relates equally to both the widely used MPEG-2 and newer MPEG-4 (H.264/AVC) codec standards. (See Figure 1.)
HD-SDI is inherently a 4:2:2 interconnect, which requires that a 4:2:0 stream be upsampled and then downsampled at each digital turnaround. This can, within as little as four to five concatenated links, result in visual blurring of the color image to the point where the quality and integrity of HD content can be severely compromised.
This is of particular importance in applications such as HD digital electronic newsgathering, where the use of lower data rate encoding and 4:2:0 chroma sampling can provide operational and economic advantages. In HD-DENG, a significant number, possibly as many as 10 or more, of concatenated operations can take place before final delivery to network affiliates and others.
Chroma encoding formats
The two most widely used chroma encoding formats in broadcast TV applications are 4:2:2 and 4:2:0. (See Figure 2.)
The original HD 1080i picture shown in Figure 3 has 1920 pixels in the horizontal scan and 1080 pixels in the vertical. The components of the picture comprise the luminance, or Y component, plus the two U and V chroma components. It is commonly accepted that the human eye is less sensitive to chroma than it is to luminance, which enables chroma resolution to be reduced to less than that for luminance in the encoding process.
In a 4:2:2 format, the two chroma components are therefore downsampled by a factor of two from the horizontal luminance component, resulting in a pixel ratio of 960 × 1080; whereas in the 4:2:0 format, the U and V chroma components are downsampled by a factor of two in both the horizontal and vertical, resulting in a pixel ratio of 960 × 540.
Chroma sampling and MPEG-4
The recent introduction of the MPEG-4 (H.264/AVC) compression standard has delivered yet another improvement in digital network use. Most MPEG-4 content can be carried in less than 50 percent of the bandwidth required for similar MPEG-2-encoded content. MPEG-4 HD video performance at encoded data (bit) rates below 10Mb/s is generally accepted as sufficient to meet many broadcast TV operations, including HD-DENG.
For lower encoded video data rates, those at or below 10Mb/s in MPEG-4 (H.264), 4:2:2 chroma sampling may not produce any significant improvement in performance compared to that of 4:2:0. It also requires more bandwidth for a specific encoded data rate. Content encoded at 10Mb/s 4:2:0 requires approximately 20 percent less bandwidth than 4:2:2-encoded content.
This presents both an opportunity and a challenge: whether to use 4:2:2 chroma sampling and accept about a 20 percent increase in bandwidth cost, or to use the more efficient 4:2:0 format with potential chroma degradation in concatenated links.
Precision filter technology
One solution now available uses a precise set of filters that process the 4:2:2 chroma component and downsamples it with a low-pass filter and 2:1 vertical sampler to create a 4:2:0 output. At the receive end, the 4:2:0 chroma component is accurately upsampled to recreate a 4:2:2 signal, as shown in Figure 4, that can be passed to the next concatenated link over an HD/SD-SDI interconnect.
This process allows an encoded 4:2:0 stream to be carried efficiently over any digital communications link.
To maintain non-degraded 4:2:0 chroma resolution throughout the concatenated chain, it is necessary for all encoders and decoders in the transmission link to use identical filter sets. However, the 4:2:0 stream may be decoded with degradation, but without undue errors or color displacement by decoders not fitted with the precision filter technology. (See Figure 5.)
The precision filter technology has been proven in tests demonstrating its ability to protect the integrity of the 4:2:0 chroma component in up to 16 concatenated encode and decode operations. Now, let's examine the protection that such precision filter design provides.
In a series of tests, a comparison between the original chroma PSNR was made with that at the output of the second, fourth, eighth and 16th concatenation of the 4:2:2/4:2:0 conversions. Test results shown in Figure 6 demonstrate chroma degradation in conventional concatenated operations and the level of protection provided by the non-degraded 4:2:0 precise filter solution that maintained PSNR close to that of the original 4:2:2 source.
It is generally recognized that downsampling and upsampling is more difficult for interlace than for progressive video largely because of nonlinear phase characteristics. The generation of unwanted artifacts is also more pervasive. Even so, the precise filtering discussed here properly handles both video formats.
It appears that such a solution can enable longer-distance encoded video communications links with less degradation. As such, many of the problems of multiple concatenations could become a thing of the past.
Author's note: The original research and material of Akira Nakagawa of Fujitsu Laboratories contributed to this article.
Keith Dunford is currently a consultant for the video solutions group of Fujitsu Computer Products of America as well as managing partner of The Exam Group.
