By Werner Ruotsalainen on Sun, 07/03/2011
I've made a lot of pretty advanced resolution tests today to find out the effective (real) resolution of my iPhone 4 and 3G S hacks and enhancements.
For this, I've used the well-known ISO 12233 resolution chart available HERE for printing / ordering. Please read THIS (additional info in THIS / THIS / THIS article) in order to be able to correctly understand how the frame grabs should be evaluated.
The test setup was as follows (as usual, click the image for the original, full-sized one):
(iPhone 4 being tested; to the left to it, a HDV-Z96 LED Photo/Video Light Kit)
That is, I lit the chart with a prosumer video light to make sure none of the cameras / iPhones use higher ISO and/or avoid smearing and to use a lower aperture meaning higher depth-of-field to avoid focussing problems or inconsistencies and, in general, less soft lens characteristics particularly important with cameras with superbright [here, f1.8] lens like the Nikon P300. (Speaking of lens corner softness, which is pretty apparent in the shots of the P300, I've only paid attention to the center of the image, which doesn't suffer from any kind of softness, not even wide open.)
During this, I've noticed the iPhone 4's effective video resolution is very high in the default (HD) mode and is even better than that of the best generic Point and Shoot camera of 2011 according to TIPA (additional discussion HERE), the Nikon P300. Take a look at the following video framegrabs:
(iPhone4; original MOV video file HERE)
(P300 720p mode; original MOV video file HERE)
(P300 1080p mode, 18 Mbps [12 Mbps has exactly the same resolution characteristics]; original MOV video file HERE)
As you can see, the P300's 720p and 1080p are almost not different. The latter has a little bit better effective resolution: check out particularly the top left corner of the crops: with the 1080p shot, aliasing and line merging “kicks in” a bit later than with 720p (around unit 7). The difference, however, is well below 0.5 units. Under unit 7, the, at the top, the nine semi-parallel lines are shown as four before merging and completely disappearing at 8.5-9. The same (the lines are no longer shown as nine distinct ones) happens to the iPhone only under unit 8 and it, here, doesn't even show any aliasing artifacts.
On the right, however, there are a lot of aliasing artifacts, even worse than those of the P300. This (and the color artifacts on the left at around 11 and, on the horizontal stripe, between 11 and 12) shows the iPhone4 has a much weaker aliasing filter than the P300; this is one of the reasons it delivers far superior true (effective) resolution.
Kinda nice, isn't it? Now, I only wish the iPhone4 lens were far wider when taking video. 40 mm equivalent lens are plain too narrow for a lot of generic tasks (e.g., shooting a video indoors of more than one person); no wonder so many people like my widener hacks (of which I'll release a new version, along with a beginners' simple tutorial, VERY soon!)
I'm only disappointed because, frankly, I would have expected much better video quality from the Nikon P300. Well, it seems I'll sell it if the Sony A77 turns out to be as good as rumored and it indeed receives a bright zoom lens starting at 24 mm equiv...
EDIT (07/4 16:11CET): Photo quality
The photo quality (here: the effective resolution) of the Nikon P300 is, on the other hand, excellent and is considerably better iPhone4. The differences can easily be spot in the next full-resolution still photos (I've also included the 3G S in the test; as usual, click the images for the full version):
(the 3-Mpixel iPhone 3G S; merging around 1300-1400 lines)
(the 5-Mpixel iPhone 4; merging around 1600-1650 lines and, with this, substantially better than the 3G S)
(the 12-Mpixel Nikon P300)
Actually, the ISO 12233 chart can't be used to correctly measure the effective resolution of the P300, which can also be seen in the last crop above, which clearly shows there's still all the 9 lines visible at 2000 lines. An enhanced, but freely not downloadable version would be needed to really find out the real . (Using it, as is also shown HERE, the lines merge at around unit 22; that is, at around 2200 line pairs, below the Nyquist limit of 3000. The maximal resolution of the ISO 12233 chart is 2000 line pairs, which can be used to correctly evaluate cameras with a CCD/CMOS having, at most, 8 Mpixels.) Unfortunately, DPReview's new, high-resolution chart is only available for purchase (for several hundred US$'s) (see the second post of the DPReview team), not for download and I couldn't find any high-res charts. Time to write a Cocoa (OS X / iOS) app to generate such a chart, maybe? ;-)