iZoom, FZ200, LX7, FastStone - a few thoughts

Started Apr 10, 2013 | Discussions thread
Detail Man
Forum ProPosts: 15,112
Like?
Re: iZoom, FZ200, LX7, FastStone - a few thoughts
In reply to Mikedigi, Apr 10, 2013

Mikedigi wrote:

Detail Man wrote:

Poked around a bit looking for Panny Intelligent Zoom information. Here's a recent blog post:

http://blog.almalence.com/test-of-panasonic-lx7-intelligent-zoom/

So he finds it pretty good for what it is.

The author finds it unremarkable in its effects. The other poster's (on this thread) example showing the re-sampling artifacts of i-Zoom compared to Digital Zoom (on the FZ150) does not make me very excited about letting i-Resolution sharpening occur. It looked pretty bad to me.

Looks like one of those features that Panasonic believes "less is more" where it comes to disclosure:

http://panasonic.net/avc/lumix/compact/lx5/engine.html

Ah, with LX5 the iZoom does not double the magnification, it increases it from 3.8x to 5x?

That's what it looks like. A case where the user appears to have less control over the situation ?

In general, these cameras (I believe) all use the Bilinear re-sampling algorithm to re-sample images (which uses a 2x2 array for computations), as Bicubics, Lanczos, and other larger array algorithms would very likely be too computationally-intensive for the in-camera hardware.

Bilinear does well with up (or down) sampling ratios of exactly 2 - but not so well otherwise ...

Using "i-Zoom" with 2x sensor-crop, up-sampling by a factor of 2 sounds like the best way to use it. . . . . .

So if I use , in iZoom, any even number, say 32x (2 x 16x), or 38x (2 x 19x), that will not be as good as using 48x?

I.e. is it one-spot, or sliding scale? I've asked Detail Man whether the best result is only achieved at 48x, or whether any even number will do, e.g. 26x, 28x, and so on up to 46x. I suspect that I know the answer.

(Possibly), it does not know or care what the optical zoom factor is. It is just using 1/2 of the image-sensor (1/4 of the area) and up-sampling that, I presume. So it is narrowing the Field of View by a factor of two (or whatever) regardless of the optical zoom.

(However), I am not familiar with the specific user-interface (whether the camera allows that simplicitly - or does things like EZ Zoom does where the sensor-cropping is used in varying amounts depending upon what the Zoom Factor is set to by the user. By all means, you would want to use it at a factor of 2 (so that you don't get grody up-sampling artifatcs from the Bilinear re-sampling that goes on). Same for Digital Zoom (keeping it to a factor of 2 is going to be best).

Am not sure whether the up-sampling occurs prior to or after the in-camera JPEG engine. Perhaps after (because that would mean the JPEG engine could operate on a smaller image, less pixels, less work?). If so, I would think it better (if possible) to record and extract a RAW file of the smaller (sensor-cropped) photosite dimensions, and RAW process it. Don't know whether that or Digital Zoom will record a sensor-cropped RAW image-file. If so, such a RAW file would be free from the "i-Resolution" sharpening (which is an in-camera JPEG engine thing).

Perhaps you record JPGs only anyway. The RAW approach (might) be an additional possibility.

Also it looks as if the Resize/Resample in FastStone will do the doubling thing, so I can try that.

And it has algorithms like Lanczos-3 which are much higher quality - especially important for non-integer re-sampling ratios deviating from a factor of 2 (where the cam's Bilinear looks best).

As with EZ Zoom with my FZ8, FZ38, FZ200, I expect the main iZoom advantage will be the user one - both camera and I seeing and focussing on the target more easliy, especially when I am spot-focussig with twigs etc in the way.

Some tests were performed on an FZ50 in EZ Zoom modes that seemed to indicate that the indicated rectangular Focus Area remained roughly as indicated in the user-interface (and IIRC, the associated Metering Area, as well). If so, note that the number of photosites that is being used in the CDAF system is reduced as a result - and the AF performance may suffer as a result. So, it's nice that you can preview the focus target better - but the AF may suffer for that ...

Whatever the case, the Fcoal Length is not modified - only the Field of View is modified, and one sacrifices quite a lot of pixel-resolution (75% of the total pixel area) in order to sensor-crop by 2.

Your examples don't contain much of any high spatial-frequency (finely detailed) subject-matter. Try it on something finely detailed, and the results may well not please or be adequate. And the i-Resolution sharpening applied may well just make matters worse on top of that.

Reply   Reply with quote   Complain
Post (hide subjects)Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow