An often forgotten fact about Sir Isaac Newton is that he was an alchemist. Dictionary.com defines alchemy as:
A medieval chemical science and speculative philosophy aiming to achieve the transmutation of the base metals into gold, the discovery of a universal cure for disease, and the discovery of a means of indefinitely prolonging life;
A power or process of transforming something common into something special;
An inexplicable or mysterious transmuting.
While Newton was interested in the first definition, today we are more interested in the last two. You might find the second definition similar to the conversion of SD pictures into HDTV images, which is not at all unlike alchemy. Those “wizards” who succeed at finding ways of transmuting SD into HD are no less valuable in our technological industry than an alchemist would have been to the economy of Newton's time. The grail we seek would allow us to repurpose images acquired in a simpler time and deliver high-quality images from those golden assets.
Which leads us to the third definition. It is hard to understand how we might actually achieve such a mysterious transmutation of a low-resolution image into one with more information content. On the surface, it would be inexplicable. But in reality, there are effective techniques that take low-res images and infer deeper content that might not be obvious. In still photography, we have all seen miraculous processing of blurry images from spacecraft that end up with stunning clarity. Amateur astronomers now routinely take photos with Webcams that, when stacked with hundreds of others and processed to death, look like Hubble took them. Why not in our business as well?
There are complicating factors. One factor is the interlace nature of most image acquisition today, and another is motion (complicated by interlace). If we could use still image techniques alone, and wait for non-real-time processing, we would get those astounding transmutations. But we seek instant gratification and insist on moving images. So approximations and best effort approaches abound.
There are other limiting factors as well. Starting from an NTSC image makes upconversion pretty ineffective, like transmuting lead into tin. The bandwidth limitations on the color channels in NTSC and the fact that they are modulated into one composite channel severely limits the usefulness of NTSC for upconversion input. By definition, NTSC has unequal bandwidths of only 4.2MHz for luminance, and 1.5MHz and .5MHz for I and Q. In practice, however, both I and Q are equal in many contemporary implementations. In addition, if done poorly, the decoding process leaves the luminance with well below the color subcarrier of 3.579MHz in effective bandwidth. Trying to convert an image with barely 3MHz of content into an HD image where native luminance bandwidth is 27MHz is a real science project.
It is far better to start with a component picture. The luminance bandwidth of SMPTE 259 signals is a true 5MHz, and the color difference signals half of that.
Figure 1. Moiré is a distortion or pattern of an image caused by conflicting frequencies from two or more different resolutions overlapping each other. A common moiré video effect is seen when someone wears a cloth containing a vertical weave or pattern, which conflicts with the horizontal scan of the camera. The resulting moving moiré can be quite annoying to the viewer. Click here to see an enlarged diagram.
The information content of an SDI image is about two to three times more dense, and it also does not suffer from cross luminance and cross color artifacts, which do not help upconversion at all. (See Figure 1.)
When an SDI image is interpolated into a higher-resolution image map, the results can be quite pleasing, though clearly a native HD image will always be superior. Even the pictures converted from 720p acquisition must be interpolated when crossconverting to 1080i or 1080p format. This conversion is nearly transparent, and because consumers do not receive a full 1920 × 1080 image, the results are quite good at the consumer receiver.
A number of conversions
Consider how many conversions an image might go through. Let's assume a consumer is watching an HD NFL sports highlight show on a DBS service. Some of those images come from SD sources, some from 720p productions and some from 1080i productions. Commercials come in as letterboxed SD and 4:3. The output of the control room is 720p, so right away, there are conversions happening.
The letterboxed SD commercials are upconverted with no aspect ratio change, but because they started as less than full image height, they are effectively reduced in vertical resolution compared to 4:3 content. The 1080i (1920 × 1080) material is de-interlaced and re-sampled to 720×1280. The production contains HDV images that were re-sampled from 1080i (limited to 1440 pixels).
In homes, the image is displayed on a plasma screen with 1388 × 768 pixels, but the output of the set-top box is 1080i only. So right there, another progressive to interlace conversion and scaling is required. How many conversions? No less than four and perhaps as many as six. If analog-to-digital conversion was also required, it would leave a pretty tattered signal. The net result is that without high-quality conversion and attention to detail, modern production techniques would not be rational, or as our third definition of alchemy says, only a mysterious process would achieve usable results.
Today's composite devices
Today many of the composite devices used in transmuting signals are based on remarkable silicon implementations. This is especially prevalent in consumer equipment, which can produce credible results for all formats. Given the manufacturing cost and volume equation, it is obvious why some of the silicon has found its way into the professional realm.
As our industry moves further away from its SD past and into the HD future, content conversion will be more important. With nearly 50 years of color SD content on the shelf and a voracious appetite for reuse, you can expect the conversion to continue to improve and prices to slide downwards. Not a bad time to be an alchemist.
John Luff is the senior vice president of business development for AZCAR.
Send questions and comments to: firstname.lastname@example.org