Question about video resolutions and quality

[tygrus2000]

Hypothetical question:

If I have a video camera or digital still camera and capture content at a high resolution, say something like 1280x960 and then resize or encode the video or picture to DVD spec 720x480, does that media have more quality, than if I would have simply captured the content at 720x480 to start with?

My gut tells me that resizing (and encoding) from a high resolution source down to a lower resolution contains more pixel for pixel content than capturing at the lower resolution to start with.

Can anyone shed light on this?

Tygrus

[g_shocker182]

Good quesiton, but try finding a camera that will record that high of a resolution for under a few (thousand) grand.

[Vejita-sama]

Yes. The general rule of thumb is to capture at the highest resolution and bitrate your system supports (uncompressed if possible). Then encode to a lower bitrate/resolution. Many people, myself included, feel this generates superior results to capturing at the final resolution and/or bitrate.

[spectroelectro]

I know for a postiive fact you will ALWYS get better esults from capturing at a higher rez then squeezing down in a program like PhotoShop to a lower rez than if you capture with the lower rez to start with. I know this becuase a friend of mine used to use a 640 x 480 digital camera to capture pics for eBay and they sucked. So we used my 3 megapixel digital camera reduced the resolution in PhotoShop and the results looked a bezillion times better.
The reason why resizing downward works better, incidentally, is that PhotoShop and other image editing programs typically use bicubic algorithms to resize to a lower rez with half-pixel processing so as to void aliasing -- this avoids the jaggies. Lo rez digital cameras (video or still) typically get stuck with the jaggies in the hardware input from their CDs, since they have no half-pixel procesing or bicubic processing built in real time. So with a low rez camera you're burned at the start with no way to get rid of the visual aliasing (unlesss you want to do a two-dimensional Fourier Transform of the image and apply complex convolutions to it in Mathematica, but that's a whole 'nuther kettle of fish. Don't get me started on Fourier Optics and video DSP, kiddies).