Clear proof of PureView's pixel oversampling
Last Saturday I made some nice shots of the Lamborghini Aventador - if you missed that post, check here. It was an hommage to the gorgeous Lamborghini Aventador I will never be able to afford, and to the still amazing camera quality of the Nokia 808 PureView.
But it also gives me the opportunity to closely watch the differences of a full resolution picture, and one made using PureView's oversampling technique - if you are not sure what that is, you'll find an explanation here. I made several pictures twice, once in full resolution (hence 34MP in 16:9) and once in PureView's 8MP resolution.
Zooming in closely on the details of the white Lamborghini, you will easily notice the difference. Original formats are on the PureViewClub set on Flickr, I will link each orginal above the resized picture.
Lamborghini example 1 - resized to 640 x 360 from 34MP - original here
Lamborghini example 1 - resized to 640 x 360 from 8MP - original here
Yes, a slightly different angle, but despite the fact you don't see anyone in the pictures, I wasn't alone there, had to make way for someone sometimes.
Detail cropped to 640 x 360 from 34 MP original
Detail cropped to 640 x 360 from 8MP Original
Lamborghini example 2 - resized to 640 x 360 from 34MP - original here.
Lamborghini example 2 - resized to 640 x 360 from 8MP - original here
Detail 1 (left top corner) cropped to 640 x 360 from 34 MP original
Detail 1 (left top corner) cropped to 640 x 360 from 8MP original
Detail 2 (center) cropped to 640 x 360 from 34MP
Detail 2 (left top corner) cropped to 640 x 360 from 8MP original
You can see the difference in "noise" quite clear for yourself, and if you'd like to see more, feel free to check the original files linked above. Also all other pictures made at Lamborghini in Leusden (the Netherlands) can be seen on the PureViewClub set on Flickr.
Update: thanks for all reactions (below). I am sure you can work miracles with the computer, the only thing I wanted to show is the 808 PureView can now do this computing by itself... I simply copied several parts (640 x 360 pixels) out of both original pictures. I think that shows the difference between these examples, and I don't understand how this test would be "misleading", but I might be missing a point here. Please correct me if I'm wrong.