iPhone and camera phone specialist, Werner Ruotsalainen, left this as a comment in Ali’s 1020 unboxing. Apple has made many claims about their 5S having a larger sensor (whilst dissing more MP), taking better photos, having a specialised LED flash and Werner has had a look at DPReview’s studio shots of the 5S to see whether there’s substance behind the Apple hype (so far, user tests have shown 1020 is still superior to the new iPhone 5S.). The 5S isn’t a bad camera at all. It’s just important to point out that where camera is concerned, the iPhone is not the end all and be all that some think it to be camera wise.
Werner makes an interesting comment on the 1020′s oversampling. Although this has been praised by reviewers for combining several pixels to make a super pixel, he notes that this can still be improved as evident supposedly by ‘oversampling’ the larger, full sized pictures manually.
DPReview has just published their standard studio shots of the iPhone 5s:
The same database also contains the 1020, the iPhone 5 and some other smartphone flagships.
A quick note: Regrettably, the Nokia 808, the LG G2 or the Sony Z1 aren’t in the DPReview database. Regarding the latter, only the older Z is available and it delivers really inferior results to any of the current flagships. As far as the current data on the LG G2 is concerned, it, regrettably, has considerably lower video resolution and higher sharpening than the rest of high-end phones (1020, iPhone 5/5s, GS4). It has around 900p resolution according to GSMArena’s comparometer athttp://www.gsmarena.com/vidcmp.php3?idType=3&idPhone1=5543&idPhone2=5371&idPhone3=4910 , while the iPhone 4S and 5 nicely deliver true 1080p. As I’ve pointed out several times, the Samsung GS4 doesn’t seem to have a low-pass filter. The results of this can easily be spotted in the video results: as with the faulty downsampler of the 1020, it introduces tons of false detail over the Nyquist threshold. This is unacceptable. Let me the real-world results of this, compared to the 808 and the LG G2 (I’ve selected the G2 here to prove it indeed has soemwhat lower resolution in video than the 808 / all the 1080p-capable iPhones):
See the absolutely awful aliasing effects over the 10-11 signs, not present in the 808 / G2 framegrabs?
The 1020 has, when using its full-res mode and doing the downsampling on the desktop (to avoid the messed-up in-camera downsampling in the 1020), has considerably better IQ than the iPhone 5s, let alone the iPhone 5. (The 5s indeed has considerably less noise but it’s in no way as clean as the 1020, assuming the same target size. You can easily make the DPReview comparometer do the downsampling online by clicking the “Print” icon in the upper right.)
I’ve made two crops clearly showing this. The first ISO12233 reschart crop + color saturation tester:
As you can see, the 5s, while having the same effective resolution, produces a far less noisy image than the iPhone 5. The properly (again, not using the currently pretty cr@ppy in-camera downsampler!) downsampled full-res image of the 1020, on the other hand, delivers way less noise and considerably more effective resolution. The GS4′s downsampled image has about the same noise and resolution as the 5s (and significantly less noise than the 5). It, therefore, can’t come close to that of the 1020.
A “real-world” (read: no-12233) resolution and noise tester:
The 1020 is the best of the four here too. The GS4 shows a lot of oversharpening effects but, despite those, just can’t deliver the “Nationalbank” subtitle reliably. The subtitle is almost unreadable on the two iPhones’ shot. The rendition quality of the upper right coin, with pretty much homogenous color, is the best on the 1020. The upper left one’s rendition particularly suffers from the GS4′s oversharpening.
Corner softness & CA in practice
We all know how bad the 1020 is in the corners (as opposed to the 808). Let’s compare it to the three above smartphones!
(Upper right corner. The lens behave in exactly the same way on all the four tested cameras in all corners. That is, this corner is representative of corner sharpness & CA.)
As you can see, the 1020 definitely lags behind the others. Yes, as I’ve recommended in my previous article athttp://mynokiablog.com/2013/07/18/mnb-rg-nokia-1020-still-image-quality-any-good/ , you WILL want to use cropping if you shoot in 16:9. Let me point out again that, if you crop 16:9 to 3:2 horizontally but not vertically (that is, not removing any of the pixel rows, only the columns), you can get rid of the entire offending area, which is 10% on both sides. Just make sure you compose your shoot accordingly and don’t let anything important in the outermost 10% region. While this certainly reduces the field-of-view of the lens, it’ll still be around 33mm equiv after the crop – that is, that of the iPhone 5 / 5s in stills mode.
Interestingly, the 5s seems to behave somewhat worse in the corners than the 5. The latter is definitely sharper there. The best is the GS4, oversharpening aside.
All in all,
- the 5s delivers better, less noisy images than the iPhone 5
- it, however, has somewhat worse corner sharpness than the previous model (as was easy to predict, given that the lens is brighter and the sensor larger, while the lens is of the same size. 808 vs. 1020 effect, albeit not as pronounced.)
- the 1020 delivers considerably better images, both noise- and resolution-wise, than any of the iPhones, if you shoot full-size and downsample on the desktop. Before it’s fixed, avoid using the output in-camera downsampler for serious shooting!
- however, the 1020 has definitely worse corner & left/right border sharpness than even the iPhone 5s, let alone the, in this regard, better-than-5s iPhone 5. Shoot in 16:9 and crop afterwards to completely get rid of the problem.