Read 30MP webcam

Real 30MP. My bad.

Interesting.. I've never noticed 'M' used to represent the wan4 character.

...also fascinating that wan4 is supposed to be a simplification of the Buddhist swastika symbol..

I am absolutely certain the image sensor is not 6400x4800. This is like those "4GB" flash disks that are 2GB disks with a fake partition table (and unusable second half).

Scanners sometimes have stupidly high resolution images available that use interpolation to generate enormous files from very little information.. like making a 4800 or even 9600 dpi file from a 1200 dpi optical resolution sensor. 16x/64x the file size. Maybe that's what they are doing. It's silly, in any case.

I just bought two of that same camera, but I paid Amazon $7 for each one. What a ripoff!

It's not bad. The lens has adjustable focus, so I can set it at infinity for my cabin-cam; most cheap "webcams" are fixed-focus at some silly compromise distance.

The USB cord has an inline bump with a pot for adjusting the LED illumination level. The LEDs are white and don't seem to be very effective at distance. Heck, for $7 you can use it as a USB adjustable illuminator and ignore the camera function; a powered hub and a few of these could make for interesting lighting.

Nice base and mounting clamp, plus a female thread for tripod mounting. The pushbutton on top tells Windows to snapshot and save a jpeg.

It doesn't come with software, and the "manual" is a tiny piece of paper with printing on one side. You can plug it in and XP will run it, display realtime and save snaps, without loading drivers.

How in the world can anybody sell this for $6? People would have paid $6000 not that many years ago.

John

On a sunny day (Tue, 8 Feb 2011 07:42:47 -0800 (PST)) it happened linnix wrote in :

Read the site, it says: Static pixels: Can capture 6400*4800 (30.0M) It wont be 30fps at that resolution. USB2 is probably not fast enough, not even in MJPEG. On the lens it is written '10x digital zoom', this confirms the picture size versus the pixels.

There's a factor of four for Marketing Megapixels--it takes four actual monochrome pixels to make one colour pixel, but of course they always quote the higher value.

Cheers

Phil Hobbs

Seems fair enough, given the way that eyes work. There are about 1.5E6 color elements of each of the three types, vs 90E6 brightness measuring elements, which fits fairly well with the NTSC ratio of luminance to chroma bandwidths.

CRT monitor pixels are specified as three colour dots = 1 pixel. So are colour LCDs. So are bitmaps. It's just cameras that have Marketing Megapixels.

What's more, there are crossed walkoff plates in front of the sensor, to smear each pixel out across all four. That helps reduce moire patterns, but it certainly reduces resolution to a quarter of the MM value.

Cheers

Phil Hobbs

Not that I disagree with you about the point here, Phil, but just to be contrary... I *have* seen some cheap no-name QVGA LCDs advertised as being "960x240" :-)

re

That would be too obvious.

Many VGA LCD/cameras are marketed as 640x480, but a repeated pattern of

R G B Y

in 320x240 cells.

Wow, those camera guys get around!

Cheers

Phil Hobbs

Good point, although I am aware of at least a couple of LCD counter-examples (from the camera world, natch)- the LCDs on both one of my DSLRs and one of my point-and-shoot cameras are said to be "920,000 pixels", which is VGA * 3.

Interesting.. so a 2560 x 1600 resolution monitor is about exactly right to show the entire image from a 12MP camera.

Kind of an optical low-pass anti-aliasing filter?

Exactly. Aliasing is a time domain moire pattern.

Cheers

Phil Hobbs

Rather than a physical system that must be aligned correctly, wouldn't it be better to handle this in software? Spatial frequencies appear in the fourier domain and seem easily dealt with. Just asking why go to the trouble aligning crossed walkoff plates?

Also, if one could take enough specific pictures it seems possible to develop, via inverse fourier, the convolution tha the walkoff plates represent and roughly get back the fuller resolution... with noise added of course?

I'm glad you mentioned this, though. I hadn't been aware.

Jon

linnix skrev 2011-02-08 16:42:

M = 1000 in the Roman numbering system. This was available way before the ISO standards.

Who knows what the Chinese invented before that.

Once it's aliased, you can't get rid of it. That's the whole problem.

Cheers

Phil Hobbs

Ah. I had missed that point about what you'd written. Thanks. Now I get it.

Jon

Yeah, anti-moire software would turn a zebra into a horse.

John