FEEDBACK
May 1, 2007 12:00 PM
Clarifying Kell factor
John Luff:
In the December 2006 issue of Broadcast Engineering, you respond to a letter that raises a question about the relation between the Kell factor and resolution. You state, incorrectly, that there is a difference in how the Kell factor applies to interlaced scan and progressive scan.
I have attached a short article that I wrote in response to a discussion about NTSC resolution that appeared in the Letters to the Editor section of the IEEE “Consumer Electronics Society Newsletter” in 2005. My response was included in Letters to the Editor in the Winter 2006 issue.
Bernard J. Lechner
John Luff responds:
My response in the December 2006 issue may have led one to believe that it is a mathematical relationship, which would be incorrect. I also incorrectly stated that the Kell factor applies to interlace and not progressive images. In fact, it only applies to the perceived resolution of a scanned image in the vertical dimension as a percentage of the limiting resolution (the Nyquist limit of half the line rate in the vertical dimension). The value 0.7, which is often quoted for the Kell factor, is heavily affected by such factors as scene content, display performance and type, optical effects, and resolution of the imaging device and display. It is, however, accurate to say that the perceived vertical resolution of a progressive 720p picture is about the same as that of an interlaced 1080i picture. The relationship is affected by many parameters. There is no doubt 1080p shows more vertical resolution than 720p.
Tom Lewis from KSAZ-TV in Phoenix wrote about the same issues. Lewis said, “If I have 18Mb/s to format my DTV stream in, whether 1080i or 720p, I already have plenty of bits to fully encode HD video without added artifacts due to bit starving … and having a higher data rate … would not yield any improvement over having just enough bits for complete encoding, regardless of progressive or interlaced content.”
However, the writer misses an important point. He states incorrectly that 18Mb is sufficient to encode any HD content. While MPEG-2 improvements have made pictures much better in the last few years, 18Mb is only sufficient for some scene content, and is particularly compromised with complex content in sports. The process of encoding an interlaced picture requires more bits, perhaps because the samples in the vertical direction are not time coherent.
Consider interlaced scanning as interleaving two frames of half the vertical resolution. In the case of 1080i, think of two 540p images offset vertically so that they will perfectly interleave. Interlace was invented as a method of analog compression. It allows systems to broadcast with the apparent resolution of twice the number of actual lines in each temporal sample without doubling the amount of data in the channel. Your eye integrates the two images and does not see flicker between them if the field rate is high enough. The viewer perceives much higher than 540 lines on each full frame (two fields), though the sampled image resolution is lower on coherent samples.
Imagine a picture moving horizontally between fields of the interlaced image. The horizontal offset of the two pictures cannot be properly represented in interlaced systems, having the effect of reducing the apparent vertical resolution and showing artifacts even before the MPEG compression process begins. The constructed interlaced image has samples that, in the vertical direction, are not coherent and therefore contain spurious information that does not encode as well as natural images, hence a loss of encoding efficiency.
In “The MPEG Handbook,” (Focal Press 2001), John Watkinson states, “Interlaced signals are harder for MPEG to compress. The confusion of temporal and spatial information makes accurate motion estimation more difficult, and this reflects in a higher bit rate being required for a given quality. In short, how can a motion estimator accurately measure motion from one field to another when the differences between the fields can equally be due to motion, vertical detail or vertical aliasing?”
It should be noted that the total number of active pixels to be encoded per second is about 11 percent lower with 720p60 than 1080i30. This is not a plea for all broadcasters to use 720p60, but rather shows that 1080p or 720p would provide better natural images and compress better (in film or video content) than interlace images. The EBU caused quite a stir in 2005 when it published a paper, stating, “On emission standards, the EBU favours the use of progressively-scanned formats, such as 720p/50 or 1080p/50, rather than interlaced formats such as 1080i/25.” (To read the paper, visit www.ebu.ch/en/technical/trev/trev_301-editorial.html.)
In the opinion of some experts, 1080p50/60 compresses essentially as well as 1080i25/30. At IBC2006, David Wood of the EBU said, “We know that 1080p/50 is virtually as efficient a broadcast format as 1080i/25.”
I am interested in your thoughts. Please write to me at
John.Luff@penton.com.
| Want to use this article? Click here for options! |  |
