Hacker News Reader

Hacker News

Comment by yread | Hacker News Reader

Parent Full thread

yread

7728 x 5368 pixels on a 1/1.2" sensor

http://europe.nokia.com/PRODUCT_METADATA_0/Products/Phones/8....

I hope they will release it also with a non-dead phone OS

View on HN

with so much noise per pixel that they are forced to bin 7 pixels together to produce a final 5MP image....

It is actually a very interesting approach. The traditional way to fight noise is to have sensor elements as large as possible within the limits of sensor size and resolution.

Basically having larger sensor elements is like having a bunch of smaller elements that cover the same area and averaging them.

But if you can read these smaller sensor elements probably you can do better than just averaging in software, if you have good noise model and image model.

I am really curious to see what algorithms they use for demosaicing/downsampling, and what kind of lowpass filter is in front of the sensor.

yes, but binning pixels is inferior to averaging because the noise in adjacent pixels is correlated. (I had a nice paper about that, still looking for it)

If the noise model can take into account the neighbouring pixels correlations this is even better, because information about the noise in one pixel gives away information about noise in close by pixels.

The worst case is when the noise is uncorrelated. Think about the other extreme, when noise in all pixels has correlation 1, i.e. it's equal. Then it is extremely easy to remove.

(BTW, I would be very interested to the paper you are talking about, if you can find it :) )

but the problem is that you do not know what is signal and what is noise. If you have two pixels with correlated noise, you have in the extreme case s_1 + n and s_2 + n, so averaging that gives you (s_1 + s_2) / 2 + n. If the noise is uncorrelated, then it might be positive in one pixel and negative in the others, so it might cancel itself out. In the extreme case you would have s_1 + n_1 and s_2 + n_2 and (n_1 + n_2) / 2 < n. Thus you have reduced the noise.

The linked article says "Nokia showed me poster-sized samples captured with the 808 PureView (printed entirely unprocessed) that basically had zero noise in them."

Yes, but that is probably after the 7:1 sampling, since that is apparently what the phone outputs.

It is probably a very good camera, but together with Symbian I fail to see the point of it, the market for Symbian high-end camera phones must be minimal at best.

Unless possibly if it is a 'prototype' that they decided to sell to try and convince MS that they need to be able to do similar stuff but with Windows Phone.

so, while Nikon and Canon are still not there at 41MP with full frame sensors, Nokia did it with a tiny phone sensor? if true, they should drop making phones immediately.

I still remember the hype around the (5MP?) camera on the N95 back then. Even the Nokia published samples were horrible, but the press made it sound like that was the best camera phone ever...

Nokia's camera phones have been pretty good in picture quality. My favorite was the N900 with its Carl Zeiss lens and the FCamera drivers from Stanford's Camera 2.0 project (http://graphics.stanford.edu/projects/camera-2.0/).

Processing the pictures took forever, but the results were quite nice. Look at this one I shot for example: http://www.flickr.com/photos/bergie/5293597184/in/set-721576...

If Canon or Nikon wanted to produce a stupidly high (100MP+) megapixel full frame 35mm sensor they could. The pixel density on their compact cameras is about .3MP/square mm, a full frame 35mm sensor is 36x24mm (864 square mm), yielding a ~260MP sensor. They keep their high end bodies at a much lower resolution to increase the available dynamic range and enable shooting in much lower lighting conditions. It'll be interesting to see what they pulled off, but in low light situations (or high contrast) is where you'll see the issues of cranking up pixel density start to crop up.

It also allows more fps continuous shooting

Reply on news.ycombinator.com