Four Thirds User

Ian · #31 29th March 2012

Here are some more crops for comparison:

E-M5 base ISO, normal exposure, developed from RAW in Viewer 2 and the completed in LR4.

Sony Alpha A55 SLT recovered from -2EV under exposure from RAW in LR4.

Sony Alpha A55 SLT normal exposure developed in LR4.

The Sony Alpha 18-55 kit lens is pretty bad in the corners as you can see. It's very good in the centre of the frame though.

Ian

mige0 · #32 29th March 2012

Quote:

Originally Posted by Ian

Can I be sure what you mean - do the bricks in the RT developed crop you posted look normal or not? They look OK to me. But then they are not in a brightly lit situation which is where the problem seems to lie and they are not from an image that has been pushed by two stops.

Ian

The effect I refer to, I guess would be much clearer visible if we would have OM-D_5 versus PEN / E-5 samples with HG glass available.

As I'm not aware of any such apropriate RAW samples, I'll try to explain as good as I can:

Please have a look at the up-sampled crops below (taken from the RawTherapee crop I already did show)

a) To me the upper image is way too soft / sleek / smoothed *especially* as the color shading within fairly similar areas is concerned (the bricks that is).
Edge contrast on the other hand is ok, as far as I can tell (remember its plain "default" setting of RawTherapee).

b) The lower image does show - in contrary – that minute details got IMO overly exaggerated (presumably to counter weight too highish NR that otherwise would soften the image too much)
This is creating IMO an artificial touch overall, also with respect to those overly „highlighted“ details that appear to be non-related.

c) Please remember what I've shown regarding noise comparison with my normalized black&white crops. (in the other thread http://www.fourthirds-user.com/forum...t=10757&page=3 )
There it was evident, that overall noise of the OM-D_5 sensor was not particularly outstanding (just very slightly above GH2), the main noise „advantage“ of the new sensor lies in higher suppression of chroma noise.

So adding a + b + c for me led to the conclusion, that the „improvement“ for the OM-D_5 in (chroma) noise is basically a software fake of that particular sensor – or lets say: a simple trade off to detail reconstruction.

No free lunch here, IMO.

Again, I suggest to not take my word for it regarding that IMO sensor specific „overly polished“ image appearance – there are already enough architecture images out there to get the glimpse (certainly the above crops – showing so few pixel - are not any well suited for this)

The most interesting thing is – besides such IMO inferior images - there are also pretty and failrly well balanced pix available too, some at Robin Wong blog, some elsewhere or this one for example:

http://www.whatdigitalcamera.com/equ...html#slideshow

taken from here:
http://www.whatdigitalcamera.com/equ...le-images.html

The main difference may lie in pattern recognition routines that may act too aggressively and too unpredictable if the image provides the „right“ trigger

as an aside:
It may be worth to note that the linked portrait was IMO deliberately adjusted in colors to improve on the pale / bluish appearance of the rest of samples there (as "usually" delivered by this sensor).

Ian · #33 29th March 2012

Although there are examples where noise reduction does happen in the imaging pipeline before the camera saves the RAW file (Sony admit to this although I understand you can switch it off where it is featured) Olympus has told me in the past that no processing of the data before it is saved as a RAW file happens with their cameras. From what you are saying, you feel that the RAW files are already manipulated?

We don't know how Olympus has designed its Bayer filter for this sensor. There is a school of thought that the Panasonic sensors use two different greens in the filter and this would need to be factored in when doing the demosaicing in the RAW conversion process. Does the new sensor have the same bayer filter colour values as the earlier sensors?

I'm not impressed by what Viewer 2 is doing to E-M5 RAW files. Even with the noise filter off there seems to be some smoothing going on. I was able to get very fine details from an E-M5 RAW using Raw Therapee - much better than Viewer 2, but then had problems getting the colour right. I think you saw these on the e-group site:

Above is the original camera JPEG crop.

Above is my RAW conversion using RAW Therapee initially and then finished off in Lightroom 4 from an uncompressed TIFF.

So I feel something is not right with the processing of E-M5 raw files. This doubt will only be confirmed or otherwise when we start to see properly profiled and calibrated support for E-M5 RAW files.

Ian

mige0 · #34 29th March 2012

Quote:

Originally Posted by Ian

Although there are examples where noise reduction does happen in the imaging pipeline before the camera saves the RAW file (Sony admit to this although I understand you can switch it off where it is featured) Olympus has told me in the past that no processing of the data before it is saved as a RAW file happens with their cameras. From what you are saying, you feel that the RAW files are already manipulated?

Please have a look here :

http://www.sony.net/SonyInfo/technol...e/cmos_01.html

The conclusions I draw from that over simplified description is that actually there can be a looot going on, before a RAW file is written - which most naturally allows for a looot of "manipulation" regarding NR / image appearance.

For me its not necessarily "manipulation" in a negative way (though for the OM-D_5 I definitely don't like the outcome) but I see it more as trade offs between design choices a sensor maker *has* to take his pick from.

Quote:

Originally Posted by Ian

We don't know how Olympus has designed its Bayer filter for this sensor. There is a school of thought that the Panasonic sensors use two different greens in the filter and this would need to be factored in when doing the demosaicing in the RAW conversion process. Does the new sensor have the same Bayer filter color values as the earlier sensors?

Not sure if a different Bayer filter could be blamed.
Don't think the effects traced down so far would find a sufficient explanation this way

Also, Olympus viewer should (!) already have been supporting that in case of.
Clearly this is not happening.

Also, according to Robin Wong who did develop all (!) the pix shown in a fresh Olympus Viewer he got for that matter from Olympus Malaysia the image appearance IMO is basically the same as yours or the ones developed with RawTherapee + extensive post-processing done by others.