Among the posters responding to his article, an argument was put forth concerning the author’s semantics. George states that low bit rate HD is fake and that it’s not really high definition. One writer in particular felt that as long as the resolution of the image was 1280x720 or 1920x1080, the image is HD, regardless of bit rate. It was a futile discussion because both the author and the writer are correct, depending on the context.

From George’s point of view, if the image can’t contain the spatial information that high bit rate 720p or 1080p provides, then the images should not be called HD. (I fully agree.) The objector was only interested in the number of pixels delivered regardless of spatial content. So the objector is semantically correct, but he completely ignores the point of high definition: the delivery of as much detail and spatial information as possible.

What the objector should have appreciated is that to accommodate low bit rates, it’s necessary to low pass the video to avoid creating visible artifacts in the transfer. By low-passing the video, finely grained textures become obscured and, for want of a better phrase, the images will appear out of focus compared to images that had not been filtered. An analog to this phenomenon is a nice visual demonstration showing the differences between a 480p DVD scaled to 1080p and true 1080p with all the high frequency spatial information intact.

A kind reader sent to my attention a link that offers an interactive comparison between screencaps of the same image frame from such a scaled DVD and true 1080p from a Blu-ray Disc. An applet loads that allows you to drag a white line boundary across the image. You have to click on the vertical line to activate it. As the line moves, you can compare the two overlaid images. Perhaps the most obvious differences are in the chainmail; the true 1080p image clearly shows each and every link while the scaled DVD’s image is so soft that it’s difficult to make out any links at all. And look at the boundary between the vertical burgundy staff and the sky. The DVD’s boundary is soft and mushy; the BD’s is sharp and well defined. As I wrote, perhaps the best analogy is that the DVD’s image appears out of focus. Click here to access the image.

On a related subject (I’ll make the connection soon), my high definition disc news service provided a link to an article written by the technology columnist for a Midwestern newspaper. The writer asserts that he can’t tell the difference between DVDs scaled to 1080p and high definition. To him I’d advise an immediate trip to the ophthalmologist, but he provokes two questions. Are scaled DVDs “good enough?” And, are scaled DVDs equivalent to high definition? The clear and irrefutable answer is that they are not. Let me try to demonstrate as simply as possible.

Scalers, using various mathematical algorithms that vary in quality and effectiveness from manufacturer to manufacturer, will attempt to predict what the various 1080p pixels would have been if the original images were captured as HD; synthesized HD pixels are based on the values of the larger 480p pixels that surround them. To make clear the difficulty of synthesizing 1080p pixels, let’s look at one single 480p pixel and the 1080p pixels it contains.

Horizontally, there are 2.67 1080p pixels for each 480p pixel and, vertically, there are 2.25 1080p pixels for each 480p pixel. Let’s round that off to 3 and 2. That’s pretty close; 2.67 x 2.25 = 6.0 and there are six times more 1080p pixels than there are 480p pixels in any 1.78:1 image.


Consider the edge of a black horizontal bar against a white background; let’s simplify further by assuming that the edge falls right on the boundary between two rows of 1080p pixels. This is how the 2 x 3 array of 1080p pixels (the area of one 480p pixel) would look, excluding the sharpness of the camera lens.


But since half the 480p pixel area is illuminated by white and the other half of its area by black, the 480p pixel simply reports a single 50% gray level (ignoring gamma). Now, the interesting thing is that no matter what the original object shape is, if the equivalent of three 1080p pixels are black and the other three are white, the 480p pixel reports the same 50% gray level. So all of the following shapes (which are resolvable in 1080p) will result on the same 480p single pixel gray color:

This is highly simplified, of course. Objects are imaged at various angles; when that happens, the color of the pixel is a weighted average of illumination based on exposed area, but I digress. Here’s the big problem with scaling. Even with hints and indicators from adjacent 480p pixels, how do you mathematically decide which of those twenty possible shapes is correct. (There are actually many, many more combinations, but I’m limiting the discussion to very special cases for simplicity’s sake.) The reality is that you can’t. Scaling simply gives you a smooth representation of 480p resolution within the image space of 1080p. And, in fact, the scaled image can actually be a little softer than the original 480p image because the math can act as a modest low pass filter.

All this technobable aside, the bottom line is that you can’t get something for nothing. DVD’s 480p images contain a limited amount of information and you can’t reasonably expect that the six times greater visual information contained within a true high definition image can be deduced. So if anyone ever asserts that scaled DVD is as good as true high definition, feel free to suggest that that’s just a load of cow pies.

Now let’s get back to the consequences of low bit rate HD delivery systems. I stated that to accommodate low bit rates, the original high definition images had to be low pass filtered. So how does that affect the appearance of the image? Let’s take another very simple example, a transition from black to white on a vertical edge in a very slightly larger array of 1080p pixels.



When a video low-pass filter is applied, high frequency information is removed and the transition from black to white becomes less sharp and less abrupt; it's spread across several pixels. So the edge might look like this; it gets softer, one could almost say smeared. And the more filtering that’s applied to accommodate a lower and lower bit rate, the softer and softer the image will look. And that brings us back to the demonstration image found here. Low-pass an image enough and reduce the bit rate enough, and even with the efficiencies of advanced CODECs you’ll have an image that resembles scaled 480p or worse. All the pixels associated with 720p or 1080p will be present but the definition certainly will not be high.

I’ve often stated that less compression and higher bit rate yield better images. I’ve claimed that as a result, BD (1080p at 40 Mbps) looks better than HD DVD (1080p at 28 Mbps). So what do you think an Apple HD download in 720p resolution at 4 Mbps is going to look like? (Or, for that matter, an Xbox 360 download in 720p resolution at 6.8 Mbps?)

Is there competition between Apple and Blu-ray Disc? No. These are two entirely different markets. The former market doesn’t care about quality and the latter does. Will Apple be successful? Possibly. I can’t deny the existence of a market for those who don’t care about what a film looks like or sounds like. What a pity. But don’t you be deceived. The finest high definition experience available today and for the foreseeable future is on Blu-ray Disc.