In all my previous Nokia 808 PureView and Lumia 1020 still camera comparisons, there has been an assumption that has proved to be slightly inaccurate. You see, on many occasions a Lumia 1020 'reframed' (usually 'zoomed') shot has been compared to a partially zoomed Nokia 808 photo and the assumption has always been that there's no difference on the 1020 between zooming at capture time and zooming later. It turns out that this isn't the case, as demonstrated in today's story of mine over on All About Windows Phone.

From the feature:

It's a fair cop - there was one nuance of the way the Lumia 1020 camera works that I got wrong in my previous article (and I've now included links in that to point to this amplication). Think about the 'zoom later' facility and it's obvious that the way the 1020's Nokia Pro Camera software works is to intelligently crop/downsample the 34MP/38MP full resolution JPG as needed. So you're starting off with an initially 'compressed' version of the original image, complete with a degree of artefacts and approximations. Unsurprisingly, the result is an 'oversampled' 5MP image which isn't as pure as it could potentially have been as if it had been constructed with data taken directly from the 1020's sensor.

However, I can exclusively reveal that the engineers at Nokia had thought of this. They knew that 'reframed' images had to be compromised slightly - there was no way round this, given that an uncompressed version of the image would have run up towards a Gigabyte each(!) - so working from a 15MB 34MP JPG it had to be. And, to be fair, the algorithms used to work from this are pretty good, with admirable detail and less artefacts than you might expect. But there's another scenario allowed for.

The initial 5MP 'shareable' image isn't, it turns out, downsampled from the 34MP JPG - it's generated from the RGB data coming from the sensor chipset. It's true that the oversampling done here is being done in software rather than in hardware (as on the older Nokia 808), but there's no intermediate 34MP JPG step/file in the way. However, the moment you hit the 'reframe' tool in Nokia Pro Camera, to zoom/pan/rotate the image, however slightly, the underlying full resolution image has to be used as the new image source, meaning an extra processing step that loses some quality (however small).

In other words, having taken a photo with the Lumia 1020 in Nokia Pro Camera, whether taken full frame or 'PureView zoomed' in, your initial sight will be of the highest quality 5MP image. If you want to keep this best quality then this is the one to upload or share. Do this before you're tempted to reframe and play with the image, as this will necessitate working from the compressed original instead and will result in a slightly lower quality photo.

These differences are small, but they are demonstrable. 

I go on to show three examples, with different degrees of 'difference', according to subject type. The most dramatic is the book spine, with text and a fine 'tree' graphic, but what most AAS readers will want to see is what the Nokia 808 would have made of the exact same shot. So I did this as well.

Here's the shot, a batch of books on a bookshelf, shot in semi-macro fashion, from around 30cm (on the limit of the Nokia 808's macro focus ability, by the way):

And here are the magnified crops of the original photo from the Nokia 808 (top) and the original 5MP image from the Lumia 1020 (i.e. before any reframing had been attempted):

Look very closely and you can see that the Nokia 808 is still king in terms of naturally presented fine detail, whereas the Lumia 1020's image looks, relatively, over-processed.

However it should be noted that we're cropped in beyond 1:1, in fact magnified by an extra factor 2, and that at 1:1 the images are all but indistinguishable. Even when magnified, the 1020's result gets surprisingly close to the 808's.

As per the linked (exclusive) article though, this Lumia 1020 result is from an original 5MP image, derived in software from the RGB data coming from the 41MP sensor at capture time, i.e. before any reframing or 'zooming later', processes that revert the 1020's output to re-downsampling from a 34MP JPG file.

Whether taking full frame shots or zooming at capture time on the Lumia 1020, it seems that to stay close to Nokia 808 levels of purity, the trick is not to then fiddle with the image further using the Nokia Pro Camera reframing facility. It's a classic compromise scenario, i.e. maximum quality versus flexibility.

Comments? Does the 1020,  in this 'initial capture, no fiddling' scenario get close enough to the results from your older Nokia 808?