Megapixels and lines per inch don't tell the whole story, or the right story. The world's geometry is chaotic, and digital representations need to capture or recreate some of that chaos to look real.
For an example outside of photography, compare any state of the art video game (say, "Far Cry") on the best state of the art video card* (say, Nvidia or ATI) on a 19" LCD, with an S-VHS tape of an old TV food commercial (say, a mid-eighties salad dressing ad showing glistening water drops on lettuce and tomatoes) on a <20" TV, and which one looks more lifelike?
The analog TV resolution is comparable to 320x240 (VHS, 0.08 megapixels) or 720x480 (DVD, 0.35 megapixels) but looks far better than a game with 1600x1200 (2 megapixels). The difference is the fuzzy (organic?) edges and analog color curves. Kind of like good bokeh. :-)
Anyway, this is why I like the Nikon D2X vs the Canon 1DS MII. Nikon's smaller sensors have noise in the grey, instead of the noise Canon has in the color channels. Canon certainly has less noise, mathematically, but to my eye, the Nikon luminance noise scattered across smaller sensors looks more like film grain than the chroma noise.
Another example of this is in color printing technology for printing presses. "Lines per inch" term comes from halftone screening, which under a loupe looks like newsprint photos seen with naked eye. I worked in printing industry in late 80's, and "stochastic" printing got a lot of attention then. The idea was that instead of halftone screens with dots following a grid, you'd use a random placement of dots to better approximate the real world colors. It worked, and looked fantastic, but was very hard to control. Here's an article about that:
http://www.kpgraphics.com/white_papers/archive/stochastic.htmlThe interaction between camera and lens isn't just about math. Looking at Nikon's 4MP from a D2Hs compared to 6-8MP from other vendors, looking at Nikon's 105 portrait lens that allows you to adjust the appearance of out-of-focus objects, I think Nikon "gets it". I think they're actually trying to look less digital, not win a MP race. And if they offer new glass for the D2X or future D sensors, I'll be first in line if it's about capturing an analog world more aesthetically, not an oscilloscope metric of resolution.
* Footnote: All video cards for the past decade have offered "anti-aliasing" which manufacturers say eliminates the jaggies aka the digital look. But look closer, because the AA is only operating on edges within solid shapes (smoothing the curves approximated by polygons on a car or torso) and not between the shapes and the background (compare edge of car to background, it's still jaggy). Next generation cards are supposed to be able to AA between foreground shapes and the background. Supposedly the XBOX 360 can do that, for example. This may finally close the gap between 1950's technology TV and 2005's video games. It's not the megapixels, it's what you do with them to make them look more lifelike.
** Originally posted at DPReview.com at 11:11 PM, Friday, October 14, 2005 (GMT-5)