For example:

If that is indeed the case then there is plenty of image information to fill up a 720p signal, and it would explain why Frankie does not see any image loss. In fact the 720p signal will show you more of the scene than the Xbox was rendering.

Not all upscalers are created equal, the scaler in the 360 (along with other PC video cards) does a much better job of resizing an image than 90% of the scalers in entry level HDTVs. For an explanition on how scalers work read this article.

Consider, for example, a postscript file. Postscript code has no knowledge of the capabilities of the output device. You can render this file to a 300 dpi printer, a 600 dpi printer, or even a 72dpi device such as a monitor. To each device the file is rendered correctly. If it were not then the 600dpi ouput image would be only 1/4 the area of the same output on the 300dpi printer. Likewise the 72dpi output would be about 16 times larger than the 300 dpi output. This is not the case. 1 inch on the 300 dpi device is 1 inch on the 600 dpu device.

Clearly postscript is a different beast than Halo/Halo2 game engine rendering, but if the graphic routines within the engine were abstracted properly then one could easily manipulate the frame buffer dimensions and render the output to a larger frame, without post-processing upscaling. In fact if Bungie at one time intended to render to a larger frame (720p) and didn’t because of the hardware capabilities of the first generation XBox then I’d suspect that the frame buffer size was abstracted from the code that actually does the drawing. It is really about having two coordinate systems one for the vetor graphics and one for the physical frame buffer pixel graphics, the vetor drawing routines draw in their coordinate system and the routines that actually manipulate the pysical frame buffer would perform a coordinate conversion function (multipler) to scale from the vector coordinate system to the physical one. If the dimensions of the physical frame buffer is changed then the multiplier is changed to adapt to the new size.

My $0.02

- McGrude

Anyone know why upscaling to an image size that isn’t a whole integer factor bigger isn’t lossy? I guess I don’t understand why that would be the case.

Here’s the golden rule: anytime you edit image data, you lose the original data. However, your edit may make it look subjectively better. My claim was that editing the frame with an AA pass after Halo 2 is done with it, as it sits in a frame buffer, theoretically entails the loss of data; it may in fact look better - but this wasn’t something I could test, and I’m glad that Frankie was willing to let us know about their subjective tests.

I’ve got my Xbox 360 pre-ordered for Tuesday, and I sure can’t wait to pop Halo 2 in - it’s the first thing I’ll try. I stood in line for all night for Halo 2, and I’ll stand in line for the 360, too.

First of all, at what resolution is the 3D scene being drawn? 480p or 720p? I think it is fair to say that it is indeed wrong to call upscaling ‘lossy’, as there isn’t anything being lost. But if the scene is being drawn at 480p then upscaled the image is fundamentally having extra data created that wasn’t in the original scene, probably based on nearest neighbour interpolation. Again, this isn’t losing anything, but there is a massive difference between an image drawn at XxY and an image created at PxQ and resized to XxY.

Secondly, what antialiasing method is being used? Is it multisampling or supersampling? It would seem likely it is multisampling applied prior to upscaling, but it could be either.

Ultimately I think that the premise of this article - that this method of anti-aliasing will harm the image quality - is in certain instances wrong, and in that I can understand why you guys would be upset at that assertion. The only circumstances where I feel there is potential for a worse quality result is if this upscaling happens only to be downscaled again to show on 480p (or less - pity us poor Brits and our crappy TVs). In that circumstance Sandstone is right - there would be loss in image quality, as you’ll be throwing away data that was created in order to fit a bigger picture on a smaller screen. Think what happens to an image in Photoshop if you resize it up and down a couple of times. I can’t say I’ve heard any comments to the effect that this feature would be optional or not, but if it isn’t then I can’t imagine that lower resolution displays would end up with a better quality picture than the Xbox original.

Either way, we’ll have to wait and see, like you say, then decide.