Reading mode: Light Dark

Silkypix DOES Use Both EXR Raw Frames (4 images)

Go to last post
Here is an image comparison that gives an idea of how much - or rather how little - of the longer exposed half is used by EX/SP. The hand/light is moving downwards.

JPG:





EX/SP:





And while looking at the squirrel's food in the JPG version I began to question my "JPG is simple blending after curves" theory and sent several of my test images through EX/SP. Seems like I have to renounce that theory.

JPGs don't seem to make use of all the information in the shorter exposed half. It may even be that JPGs/other raw converters work quite exactly opposite to how EX/SP works: with priority on longer exposed half, but also using only part of the information of the shorter exposed half.

Here is another EX/SP file that contrary to other test shots shows that at least part of the light-trail of the longer exposed half is used by EX/SP. Much more interesting is the position of my hand over the b/w picture on the wall, though.





In the corresponding out-of-camera JPG there is not the faintest sign of the "hand over picture" information being incorporated/blended into the final image.





So obviously I misinterpreted some of my former data. But what disturbs me is that still clipped highlight information is blended into the final image, while unclipped and possibly "useful" information is not. Quite counterproductive for a feature that is meant to protect from clipped highlights.

And while I find all this investigating interesting I also look very much forwards to replace the EXR X10 with a camera that comes with better dynamic range to begin with and doesn't leave its user wondering about reproducible results with all its quirks and complexities.
 
Trevor G said:
There is a second arm image partly imposed upon his arm, but further back and at an angle, and partly showing up on his trousers and vest, but here, the shadow which existed from the second frame partly obscures some of that highlight detail added from the first frame.



Click to expand...
I forgot to mention that the most significant evidence that this RAW output image is a composite of 2 or more images is the groom's collar - note the bright white part of it which stands out, much like the bright, second arm hanging at his side.

Compare it with the bright, white parts on the JPEG, which are different again.

--
Cheers ;-)

Trevor G

Silkypix tutorials at: http://photo.computerwyse.com
 
Timur Born said:
So obviously I misinterpreted some of my former data. But what disturbs me is that still clipped highlight information is blended into the final image, while unclipped and possibly "useful" information is not. Quite counterproductive for a feature that is meant to protect from clipped highlights.
Click to expand...
Thanks for the admission. Where is the EXIF/shooting details?

I don't think DR is really meant for these circumstances.

It's really meant for bright, sunny days. Then it works exactly as planned, giving 2.7EV of highlight correction, or more, in RAW.

JPEGs get 1 to 1.5EV at most - the rest is compressed.

If you don't need highlight recovery, don't use DR400. If you are shooting at long exposures, use DR100 - you then won't get 2 mis-timed images with poorly overlapping material.

--
Cheers ;-)

Trevor G

Silkypix tutorials at: http://photo.computerwyse.com
 
All of which could be said about your example, too. The collar looks like it does because SP choose to blend the bright highlight details of both image halves as far as the groom's shirt is concerned.

But is the outcome useful? And do we want a priority on the shorter and thus worse exposed image half with only very slight information from the longer and better exposed half being used. In this case it helped against your wrong decision of using a too long shutter time, but that's not exactly a "feature".

Also there is no more than two images. What you see is an amalgam of using only one image in some parts, and two images in others. Where it appears like three images is where objects on the longer exposed half show motion blur, part of which is removed by SP completely, part of which is blended. Where it is removed completely you usually can see remaining artifacts, look my "red" demonstration image and look at the lighter gray remains on the darker grey trousers and look at the stones behind the back of the bride.

Here is the EXIF you asked for, but I doubt that it contains anything useful. :P

Aperture : 2.0
Aperture Value : 2.0
Auto Bracketing : Off
Bits Per Sample : 8
Blur Warning : Blur Warning
Brightness Value : -0.21
Camera Model Name : X10
Categories Icon Resource :
Categories Name :
Categories Name Translation Mode: 0
Categories Smart Search Support : 0
Categories UI Order : 0
Categories UI Visibility : 0
Circle Of Confusion : 0.008 mm
Color Components : 3
Color Space : sRGB
Components Configuration : Y, Cb, Cr, -
Compressed Bits Per Pixel : 4
Compression : JPEG (old-style)
Copyright :
Create Date : 2012:01:08 21:17:51
Custom Rendered : Normal
Date/Time Original : 2012:01:08 21:17:51
Development Dynamic Range : 400
Directory : D: Pictures/2012-01-08
Dynamic Range : Standard
Dynamic Range Setting : Manual
Encoding Process : Baseline DCT, Huffman coding
Exif Byte Order : Little-endian (Intel, II)
Exif Image Height : 2112
Exif Image Width : 2816
Exif Version : 0230
ExifTool Version Number : 8.76
Exposure Compensation : 0
Exposure Mode : Auto
Exposure Program : Aperture-priority AE
Exposure Time : 1/7
Exposure Warning : Good
F Number : 2.0
Faces Detected : 0
Field Of View : 65.5 deg
File Modification Date/Time : 2012:01:08 22:20:38+01:00
File Name : DSCF6048.JPG
File Permissions : rw-rw-rw-
File Size : 2.5 MB
File Source : Digital Camera
File Type : JPEG
Film Mode : F0/Standard
Flash : Off, Did not fire
Flash Exposure Comp : 0
Flashpix Version : 0100
Focal Length : 7.1 mm (35 mm equivalent: 28.0 mm)
Focal Length : 7.1 mm
Focal Plane Resolution Unit : cm
Focal Plane X Resolution : 3205
Focal Plane Y Resolution : 3205
Focus Mode : Auto
Focus Pixel : 1408 1056
Focus Warning : Good
Fuji Flash Mode : Off
High ISO Noise Reduction : Weak
Hyperfocal Distance : 3.30 m
ISO : 100
Image Height : 2112
Image Size : 2816x2112
Image Width : 2816

Internal Serial Number : FPX 20932001 593130323233 2011:10:26 F6E030110E5D
Interoperability Index : R98 - DCF basic file (sRGB)
Interoperability Version : 0100
Light Source : Unknown
Light Value : 4.8
MIME Type : image/jpeg
Macro : Off
Make : FUJIFILM
Max Aperture Value : 2.0
Metering Mode : Spot
Modify Date : 2012:01:08 21:17:51
Noise Reduction : n/a
Orientation : Horizontal (normal)
Picture Mode : Aperture-priority AE
PrintIM Version : 0250
Quality : FINE
Rating : 0
Resolution Unit : inches
Saturation : Normal
Scale Factor To 35 mm Equivalent: 3.9
Scene Capture Type : Standard
Scene Type : Directly photographed
Sensing Method : One-chip color area
Sensitivity Type : Standard Output Sensitivity
Sequence Number : 0
Sharpness : Normal
Shutter Speed : 1/7
Shutter Speed Value : 1/7
Slow Sync : Off
Software : Digital Camera X10 Ver1.02
Subject Distance Range : Unknown

Thumbnail Image : (Binary data 10618 bytes, use -b option to extract)
Thumbnail Length : 10618
Thumbnail Offset : 1908
Version : 0130
White Balance : Auto
White Balance Fine Tune : Red +0, Blue +0
X Resolution : 72
XMP Toolkit : XMP Core 4.1.1-Exiv2
Y Cb Cr Positioning : Co-sited
Y Cb Cr Sub Sampling : YCbCr4:2:2 (2 1)
Y Resolution : 72
 
Trevor G said:
While you chew that one over I will also post the two images extracted from the RAW via RawDigger, but that will have to be tomorrow. ;-)
Click to expand...
As promised, here are the two images extracted from the 19MB RAW by RawDigger.

Note that frame 2, the shorter exposed one, has less information than the resulting RAW which is based on it. Look at the groom's right sleeve.

The exposure difference in the lowlights betweeen these two images is just 1EV - that also means something, since they were shot at DR400. I think that the shorter exposure (frame 2) has its lowlights lifted, while its mid tones stay the same. I'll post the histograms to show...









Finally, please compare frame 2 with the RAW jpeg - you will see the differences.

--
Cheers ;-)

Trevor G

Silkypix tutorials at: http://photo.computerwyse.com
 
And to specifically answer this statement:
Trevor G said:
I don't think DR is really meant for these circumstances.
Click to expand...
Here is what DR does to clipped and blooming (up to orb level) highlights on non-moving images. Halo anyone (aka longer exposed clipping blended into the final image)?





Again I am puzzled how clipped highlights are blended while other information is not?! I would very much like to rely on AUTO DR to have DR ready when needed and off when not, but since AUTO DR is bugged and thus useless... too bad.
 
And here are the histograms superimposed for clarity.

There's much more to DR than is obvious at first glance.

Firstly, even though this is DR400 and should show a 2EV difference in exposure, the lowlights are only 1EV apart. That means something else is going on.

Further examination of the histogram shows what it is. Since the midpoint in both images is the same (it corresponds with the darker front edge to the bride's dress and is represented by the little hump ^ at the halfway mark) that means that curves-type manipulation has taken place on the lowlights, most probably to lift them on the shorter exposure.









Remember, this is ISO200, DR400.

And this is where I'll have to leave it for tonight. ;-)

--
Cheers ;-)

Trevor G

Silkypix tutorials at: http://photo.computerwyse.com
 
And the best of the three presented images is? I'd say the underexposed single half that shows no motion blur and no "we don't know why SP is doing what it's doing" raw conversion artifacts.

With both sensor halves at hand you can now try to blend both images together manually to recover some shadow to mid-tone information from the longer exposed half on the static background.
 
VictorZ said:
I don't know, but it looks to me (based on the images posted, and especially eduT's animated GIF) that SilkyPix is using the shorter exposed image for all the mid/shadow portions of the image. If so, why? That seems backwards. Wouldn't you want the brighter (and presumably less noisy) shadow/mids combined with the detail from the underexposed highlights?
Click to expand...
Yes, exactly. In the same way that the contribution from 2 or 3 drivers is combined in an audio speaker using an analog tone curve, the visuals from the two sensors should be combined on a curve.

The curve should start the brighter of the two frames at -2EV with the lowest values, and for values above some threshold should smoothly curve down to -12 EV or whatever makes the wasted bright values go away.

Similarly, the darkest frame should start at +0ev for the darker frame for the brightest values and curve down quickly as values drop below some other threshold to eliminate the contribution of the blocked shadows.

Thus, you get the least noisy tones with a flat overall tone curve. Then you can adjust in a second step, which of course would be the application of the gamma tone curve that takes the linear data into the jpeg world.

In RAW, we would see a very flat set of tones and adjust from there using the same methodology but with vastly more control. The latest LR and CS both seem to handle EXR RAW files better. I notice them jumping into that uber-bright display and then coming back to a midtone display, which tells me that Adobe has worked on this a bit.

So I agree ... I am not quite convinced yet by the evidence presented. I agree with the premise -- that RAW converters try to perform this blend -- but not that the evidence presented confirms or denies it.

--
http://kimletkeman.blogspot.com
 
Trevor G said:
edu T said:
(Only with their colors reduced to a GIF palette, not adjusted for the known scaling mismatch between the JPGs and RAWs.)
Click to expand...
What "scaling mismatch" are you referring to?
Click to expand...
Ah yes. To this small but noticeable zooming in/zooming out -like difference:

 
This scaling mismatch:

10
 
Kim Letkeman said:
In RAW, we would see a very flat set of tones and adjust from there using the same methodology but with vastly more control. The latest LR and CS both seem to handle EXR RAW files better. I notice them jumping into that uber-bright display and then coming back to a midtone display, which tells me that Adobe has worked on this a bit.
Click to expand...
Capture One does it best at the moment. It is the only raw converter that combines three important features for EXR DR raws:
  • No dark outline around the shorter exposed half's highlights.
  • Can retrieve information from the blended highlights of the longer exposed half.
  • Highlights of the longer exposed half retain their neutral tint.
I just checked ACR 7 beta and it introduces the same purple tint on longer exposed highlights that LR 4 does. Hopefully Eric Chan will have time to look into the raw files I send him upon request, but by then I will likely already be using a Bayer pattern camera.
 
Silkypix and Lightroom keep squares square, JPG turns squares into rectangles by slightly compressing the image horizontally (and thus keeping more in frame).
 
You must log in or register to reply here.

Keyboard shortcuts

f
Forum
m
My threads
Mobile site
About
Editorial content
Cameras & Lenses
Community
All content, design, and layout are Copyright © 1998–2025 Digital Photography Review All Rights Reserved.
Reproduction in whole or part in any form or medium without specific written permission is prohibited.
When you use DPReview links to buy products, the site may earn a commission.
©GPS Media - Guides, Products, Services.
Back
Top