Now why would you think that a sensor is oversampling bad data? The reality is that an increase in megapixels will, for all but the very worst lenses, increase real picture resolution. We are very far from any optical limitation on the lenses.
This isn't really true at all. I use a modest resolution (24MP) D3X and many of my lenses (still the best Nikon makes at each focal length) are utterly incapable of rendering an image that approaches sharpness at the pixel level ... while some of the newest can, when properly stopped down.
Notice how most of Nikon's example pictures for the D800 are shot at f/8? One at f/10 and another at f/4.5 (which is obviously soft by the way), and they used the best lenses they have. Must be fun if images shot far from f/8 can't be used to demonstrate the capability of the camera without embarrassment to the manufacturer, huh? A nice, 2700€ fixed aperture camera. Cool.
Take the anti-aliasing filter as an example. It is not there because the sensors have too high resolutions; it is there because they are too low res. (And the lens resolution needs to be blurred).
No. The image projected by the image from some lenses at
some apertures , some distances, and at the very center of the image circle needs to be blurred because it contains too high frequencies (these would affect very small parts of the image because most will be out of focus). But if you shoot wide open or stopped down there is no such problem. A 100MP sensor would produce an image with very little more information than a 24MP sensor when you don't use optimal aperture, optimal technique, optimal (manual live view) focus and a great lens. If your subject moves you'll have to use an extremely high shutter speed to make the extra pixels useful. What it would do is make 175MB raw files which would convert into 600MB TIFFs and if you add say, three layers you get a 1.8 GB image in the memory of the image editor, which typically requires three times that, to actually run, which is 5.4 GB just to open and edit one image. Sounds like fun? Wait until you have to write that data after editing into a file over a USB 2.0 connection. Talk about costly imaging.
Also remember that the depth of field gets more and more shallow and pixel-level noise more and more obvious when you start demanding sharpness at the pixel level with those high resolution sensors. There is an ever narrowing opportunitiy to actually make good of those pixels. But people will find out soon enough as they try to put those cameras to good use.
(By the way I am not saying that the image gets worse as you increase sensor MTF. Obviously not, except at the highest ISO. But what happens is that the improvements are so insignificant that no one really cares. That's why Canon made a 22MP camera - they made 40MP sensors quite some time ago but didn't bring them to market because they didn't think the market demand exists.)
And those are real quality enhancing pixels too: I did a test once with my Ricoh GRIII snapshot and my Canon 5D II, comparing the Ricoh image with the part of the Canon image from an equal size area. Given that the Canon sensor is about 22 times larger that is a 10 megapixel Ricoh image versus a 1 megapixel Canon image. The detail level and quality of the Ricoh is way higher - even after downsampling to 1 megapixel. So there is still a long way to go before optics are the limit.
Yes, in a tiny patch of image at the very center of a lens optimized to image a tiny image circle which is devoid of optical aberrations because it's at the very center. If you had mounted your Canon lens on the Ricoh camera and shifted it off center the picture would have been blurred.
).
