Capture One on the A7RII giving me disappointing results

[Robert Ardill]

It is possible to start with the linear curve and replicate the behaviour of the Input Film Curve in the curves tool. Vice versa it also possible to start with a Film Curve and replicate the behaviour of the Input Linear Curve in the curves tool.
Inverted curves cancel each other out when working in 16bit. Attached an image with a curve followed by the same curve inverted (prepared in Photoshop) showing that you can recover the full range of details even when you are working on a really narrow band of tonal values. Now, the Film Curve in C1 is far less steep than the curves applied in my example. According to this I don't think that you are effectively losing details ... but maybe it is more intuitive for you to start editing with the linear curve in your Westport image. As said above I also use the linear curve ... but mostly when I want to work out very fine nuanced tonal differentiation in bright image areas. The cool thing is we have the choice ...

Yes, I think you're right.  Also, from the comments above it does seem (surprisingly to me) that C1 does in fact combine adjustments and apply the net effect.  I have tried this with applying saturation and exposure and then reversing it in a local adjustment and as far as I can see the two are combined into one effect, which is very cool (although there is a color shift with saturation, presumably because of Bart's Saturation/Vibrance point).

In LR this seems to work for exposure but not saturation.  Also, there are far fewer local adjustments possible in LR.

Robert

[Robert Ardill]

This is probably nit-picking, but it seems (with C1) that if parts of the image have some channel that is blown out (at least I think that this is the issue), setting the background to exposure to +2 (say) and a local adjustment layer set to -2 with the mask selecting the whole image, that the adjustment layer negates the background adjustment, except for the blown channels.  You can see this here:



(The image shows the result of the original tif exported to Photoshop with the adjusted tif as a layer above it, set to Difference, and with a Levels adjustment layer applied).

[Robert Ardill]

Regarding white balance in C1, this is the comment from Phase One support:

"The WB difference you see simply has to do with how the WB number is determined.  In capture One we use a single scale and convert from the raw file to our scale - therefore a DNG being a different raw file with slightly different properties falls at a different place on the scale."

In other words, the differences we see in WB between DNG and ARW files is not relevant, according to P1 ... they're just numbers.  A bit odd, but there you go.

Robert

[Jack Hogan]

Regarding white balance in C1, this is the comment from Phase One support:

"The WB difference you see simply has to do with how the WB number is determined.  In capture One we use a single scale and convert from the raw file to our scale - therefore a DNG being a different raw file with slightly different properties falls at a different place on the scale."

In other words, the differences we see in WB between DNG and ARW files is not relevant, according to P1 ... they're just numbers.  A bit odd, but there you go.

I would tend to agree, it's not what the absolute numbers are but how they are used - and color temperatures correlated to a blackbody radiator maybe representative in some cases but not in others.

Incidentally, and simplifying for clarity, ideally first the converter levels the raw data (i.e. applies multipliers so that the r,g,b values on a gray card are all the same) then it multiplies the raw data by a color matrix (the result of which can further be tweaked by profiles etc. later on in the processing).  The color matrix is typically a mix of two matrices specific to lighting at two ends of the spectrum (often 2850K and 6500K).  If a gray card (or other neutral object) is used to 'sample' white balance, the only reason why CCT is needed is to interpolate between the two.  It is often calculated like so: the input to the equation are the WB multipliers expressed as xy chromaticities.

Jack

[Bart_van_der_Wolf]

This is probably nit-picking, but it seems (with C1) that if parts of the image have some channel that is blown out (at least I think that this is the issue), setting the background to exposure to +2 (say) and a local adjustment layer set to -2 with the mask selecting the whole image, that the adjustment layer negates the background adjustment, except for the blown channels.

Hi Robert,

Interesting.

What is the maximum delta of the differenced layers? Is the original really clipped there in Raw?

I wonder what happens if you leave the background layer as it is, and add two adjustment layers, one +2 and one -2, with full (inverted default) mask. Maybe the adjustment on the background layer dominates and totally clipped info is not (fully) recoverable, but layer adjustments are aggregated by control type before being applied and are therefore fully reversible.
Another approach worth testing is a -2 exposure background layer (as if focusing on highlight protection) with a single +2 adjustment layer, so the other way around from your +2 with -2 adjustment scenario.

Single channel clipping is often aggravated by White balancing, so maybe a modified Background WB (with compensation in an adjustment layer with full mask and erased highlights) will also reduce the differences.

I'm tempted to consider it a bug (from the point of reversible parametric editing with adjustment layers), but maybe it's temporarily solvable by only using masks, for those cases that need totally clipped highlight control. Once it's clear what is going on, I'm sure that Phase One support is interested in being made aware.

Cheers,
Bart

[tho_mas]

Yes, I think you're right.  Also, from the comments above it does seem (surprisingly to me) that C1 does in fact combine adjustments and apply the net effect.  I have tried this with applying saturation and exposure and then reversing it in a local adjustment and as far as I can see the two are combined into one effect, which is very cool (although there is a color shift with saturation, presumably because of Bart's Saturation/Vibrance point).

In LR this seems to work for exposure but not saturation.  Also, there are far fewer local adjustments possible in LR.

So it seems it's about time to change the title of this thread into: "Capture One on the A7RII giving me pleasant results" :-))

[Robert Ardill]

So it seems it's about time to change the title of this thread into: "Capture One on the A7RII giving me pleasant results" :-))

Yes, I think it is going in that direction  .

BTW, I have involved Phase One in this 'issue' and this is what they have said:

"Thanks for making the files available.. what I am currently missing though is exactly what it is you believe that Capture One is not doing well?   For instance I can see there are a few finer details in the stone in Lightroom, however this really realities to base level, un-removable sharpening in the rendering. At the same time Capture One appears to handle tree detail better and shows much less noise in the lightpost.*

It's important to keep in mind that certain elements of image rendering are set as default for a reason in software, and the true judge of a software is what happens when you optimize an image, rather than what happens when you turn off certain features**.  Bear in mind that things like noise reduction, sharpening, color and gamma are established and edited very differently by different software, and when we program, we program based on expected end use, rather than what an item looks like with adjustments turned off.  "Turned off" after all can mean two very different things in different software.".

Which means that in the demosaicing or rendering that C1 applies some base level sharpening and noise reduction.  Presumable so does LR.  So it would seem that it is pretty well impossible to compare like with like by completely turning off sharpening and noise reduction.  This would also explain why the 1DSIII images are clearly sharper in C1 with sharpening and noise off, whereas they are not for the A7RII (or at least not consistently so, presumably depending on the frequency of the detail).

One for thing for sure though, and that is that Phase One support appears knowledgeable and responsive.

Cheers

Robert

* not sure I totally agree with this, but I would have to go back to look at the image again.
** which is not really applicable if one does the sharpening and noise reduction on the TIFF.

[Robert Ardill]

What is the maximum delta of the differenced layers? Is the original really clipped there in Raw?

The differences are very very small but I don't know how to compute the maximum delta of the differenced layers.  If you let me know how to I'll do it, otherwise here is the difference file: Difference File.   What I have done is to sample a point on one of the lines and (obviously with the Levels turned off) I'm getting a Lab value of 1,0,0 and RGB of 2,2,3 (so it's not visible without a Levels adjustment).  Erik's image, which is more blown out in the sun, has a Lab value of 1,0,0 also and RGB of 3,3,3.  With the Levels applied to it it looks like this:



C1 does show an exposure warning on the parts of the image giving problems (even with  Exposure Warning set to 255) but it's not possible to see the clipping on the histogram.

I wonder what happens if you leave the background layer as it is, and add two adjustment layers, one +2 and one -2, with full (inverted default) mask.

This works perfectly. I tried with both a linear and film curve ... both are fine.

Maybe the adjustment on the background layer dominates and totally clipped info is not (fully) recoverable, but layer adjustments are aggregated by control type before being applied and are therefore fully reversible.

Another approach worth testing is a -2 exposure background layer (as if focusing on highlight protection) with a single +2 adjustment layer, so the other way around from your +2 with -2 adjustment scenario.

This also works fine.

I'm tempted to consider it a bug (from the point of reversible parametric editing with adjustment layers), but maybe it's temporarily solvable by only using masks, for those cases that need totally clipped highlight control. Once it's clear what is going on, I'm sure that Phase One support is interested in being made aware.

It may be a bug, but it's so minor that I wonder if it's worth reporting to P1?  If it ever became a problem, which I doubt, it can be worked around by using only adjustment layers.

I guess I might report it in case it is a bug and they can fix it at their leasure.

Cheers

Robert

[Bart_van_der_Wolf]

Which means that in the demosaicing or rendering that C1 applies some base level sharpening and noise reduction.  Presumable so does LR.

Hi Robert,

I'm not so sure that it is specific sharpening or noise reduction, but rather a method of convertion that preserves the original data better without amplifying noise.

So it would seem that it is pretty well impossible to compare like with like by completely turning off sharpening and noise reduction.

Well, turning those off in LR and in C1, IMHO, does establish a baseline (including different artifacts) which then gets adjusted with controls that do a different attempt at getting less noise and preserve sharpness, or boost acutance (and in the case of deconvolution, attempts to restore signal resolution more than noise).

This would also explain why the 1DSIII images are clearly sharper in C1 with sharpening and noise off, whereas they are not for the A7RII (or at least not consistently so, presumably depending on the frequency of the detail).

My star target test does show that more resolution is extracted from the same Raw image, and that cannot be achieved by sharpening alone. In fact, C1 attempts to recreate detail finer than the Nyquist frequency (which doesn't exist, so cannot be present nor sharpened to become more visible). It is a direct result of a different Demosaicing algorithm that makes intelligent assumptions (more often right than wrong) based on undersampled Bayer CFA information. So it seems to adapt to the kind of signal it is fed, but still has to do that with the same undersampled data we feed it.

The algorithm changed when C1 version 7 was introduced, until that time LR and C1 resolved a similar amount of detail. After the change, C1 also rendered high contrast lines/edges in a less jagged way compared to C1 version 6.

One for thing for sure though, and that is that Phase One support appears knowledgeable and responsive.

It seems to have improved, but can at times also get a bit less responsive after introduction of a major new version, due to the flood of questions that such a new release generates. Let's hope they do not outsource their helpdesk to some far away low labor cost country where communication between the development team and the support people is much harder.

Cheers,
Bart

[Robert Ardill]

Hi Robert,

I'm not so sure that it is specific sharpening or noise reduction, but rather a method of convertion that preserves the original data better without amplifying noise.

Well, turning those off in LR and in C1, IMHO, does establish a baseline (including different artifacts) which then gets adjusted with controls that do a different attempt at getting less noise and preserve sharpness, or boost acutance (and in the case of deconvolution, attempts to restore signal resolution more than noise).

My star target test does show that more resolution is extracted from the same Raw image, and that cannot be achieved by sharpening alone. In fact, C1 attempts to recreate detail finer than the Nyquist frequency (which doesn't exist, so cannot be present nor sharpened to become more visible). It is a direct result of a different Demosaicing algorithm that makes intelligent assumptions (more often right than wrong) based on undersampled Bayer CFA information. So it seems to adapt to the kind of signal it is fed, but still has to do that with the same undersampled data we feed it.

The algorithm changed when C1 version 7 was introduced, until that time LR and C1 resolved a similar amount of detail. After the change, C1 also rendered high contrast lines/edges in a less jagged way compared to C1 version 6.

It seems to have improved, but can at times also get a bit less responsive after introduction of a major new version, due to the flood of questions that such a new release generates. Let's hope they do not outsource their helpdesk to some far away low labor cost country where communication between the development team and the support people is much harder.

Cheers,
Bart

Hi Bart ... yes I'm sure you must be right regarding the 'base level' sharpening/noise reduction.  I know nothing about demosaicing algorithms, but it does stand to reason that creating a single RGB value from multiple R, G and B values will involve quite complex decisions in terms of the luminance of the single value ... and therefore the detail and local contrast.

Which makes one wonder why the P1 technician says that there is a base level sharpening applied ... perhaps he doesn't know, or it's just a way of explaining it to a customer without going into technicalities ... or perhaps it's a way of explaining away the areas where C1 appears a bit weaker than does LR for the A7RII images.

A question for you: is it likely that there would be a demosaicing algorithm difference for different cameras?  The Imatest results from the tests I did showed C1 being clearly superior to LR for the 1DSIII, but not for the A7RII.  Perhaps there is an algorithmic difference for cameras that have an AA filter and those that do not?

I could test LR/C1 with your star target test ... but is it possible to print this using an inkjet printer? (I have a Canon iPF6400).

As for P1 support ... they do seem quite knowledgeable, but they aren't quick!  I wouldn't like to try to get an urgent problem fixed using email, but I suppose it would be possible to call one of their offices and ask to be put through to technical support.

BTW ... I have reported the adjustment 'issue' to P1, so it will be interesting to see what they say.

[Bart_van_der_Wolf]

The differences are very very small but I don't know how to compute the maximum delta of the differenced layers.

I usually look at the histogram in Photoshop, and drag over it till the pixel count changes or stops to change.

If you let me know how to I'll do it, otherwise here is the difference file: Difference File.   What I have done is to sample a point on one of the lines and (obviously with the Levels turned off) I'm getting a Lab value of 1,0,0 and RGB of 2,2,3 (so it's not visible without a Levels adjustment).  Erik's image, which is more blown out in the sun, has a Lab value of 1,0,0 also and RGB of 3,3,3.

That's indeed a minor difference which would need a lot of boosting to become even visible in a normal scene. But it should not be there, so one never knows if it has other side effects.

C1 does show an exposure warning on the parts of the image giving problems (even with  Exposure Warning set to 255) but it's not possible to see the clipping on the histogram.

This works perfectly. I tried with both a linear and film curve ... both are fine.

This also works fine.

Good to know.

It may be a bug, but it's so minor that I wonder if it's worth reporting to P1?  If it ever became a problem, which I doubt, it can be worked around by using only adjustment layers.

I guess I might report it in case it is a bug and they can fix it at their leasure.

I agree, it's only a very minor issue, not worth a lot of their resources to immediately hunt it down, but still something to keep in mind if one has to deal with a lot of clipped (really clipped in Raw, not after White balancing) image data.

I've been shooting a lot of shiny silverware and glossy ceramic objects lately, and have managed with an exposure adjustment layer to selectively 'tame' highlight clipping. Some clipping is okay, since it is just the mirror image of the lightsource which should Whitebalance to 'white' anyway, but it usually is nice to have a bit of the underlying design/colors in ceramics shining through.

I have tried it the other way around as well, starting with a highlight controlled background, and adjusting the overall exposure with an adjustment layer, but doesn't really make a clearly visible difference. The drawback is that some features for the background layer, are not available for adjustment layers, so it would be more than a simple change in workflow to tackle a minor issue. Also Saturation applied on the background comes out a bit different after changing exposure than the same after a correction on an adjustment layer, so I'll stick to my usual workflow, and only correct minor local issues with adjustment layers after getting a general baseline tonality in the background that is correct.

There are also other aspects of image quality that may be different between the Raw converters. I'm also quite pleased with the new resampling quality of C1 (LR was already pretty decent, unlike Photoshop). I'll see if I can produce a comparison between LR and C1 with very small amounts of rotation and or keystone correction. Very small rotations or skews usually cause a fading in and out of sharpness (like it's shown here ( figure 8 )) that can look quite bad, and orientation and distortion correction are typical operations that trigger that.

Cheers,
Bart

[Bart_van_der_Wolf]

Hi Bart ... yes I'm sure you must be right regarding the 'base level' sharpening/noise reduction.  I know nothing about demosaicing algorithms, but it does stand to reason that creating a single RGB value from multiple R, G and B values will involve quite complex decisions in terms of the luminance of the single value ... and therefore the detail and local contrast.

Yes, there is a lot of secret sauce involved (otherwise they'd all produce the same results). Of course, C1 has a legacy of having to deal with captures from sensors without aliasing filter (not because that's superior, but because there are no affordable AA-filters at those Medium Format sizes).

Which makes one wonder why the P1 technician says that there is a base level sharpening applied ... perhaps he doesn't know, or it's just a way of explaining it to a customer without going into technicalities ... or perhaps it's a way of explaining away the areas where C1 appears a bit weaker than does LR for the A7RII images.

They are probably not telling the whole story (patented stuff), or do not know how to explain in a few lines. C1 probably adapts to the AA-filter being present or not. That could be interpreted as more or less sharpening, but I'd say that's not really what happens. It's more likely some different weighting during the demosaicing, to avoid AA-artifacts on some cameras. The A7RII is probably not pushed as far as a camera with an AA-filter. That BTW will probably mean that deconvolution sharpening also works better, because there are fewer artifacts that will be amplified along with the genuinely blurred signal.

A question for you: is it likely that there would be a demosaicing algorithm difference for different cameras?  The Imatest results from the tests I did showed C1 being clearly superior to LR for the 1DSIII, but not for the A7RII.  Perhaps there is an algorithmic difference for cameras that have an AA filter and those that do not?

Yes, I think that is likely. An adjusted demosaicing weighting for cameras with and without AA-filter seem a logical approach. Aliasing artifacts are usually larger versions of the spatial frequencies that caused them. So it makes sense to avoid, or at least not amplify, them. Aliases have the largest modulation near the Nyquist frequency, and micro-contrast of the real signal is dropping to a minimum, so near the physical limit of resolution they show there ugliness 'best'. It makes sense to take special care and not make matters worse.

I could test LR/C1 with your star target test ... but is it possible to print this using an inkjet printer? (I have a Canon iPF6400).

Yes, I have a version for Canon printers (at 600 PPI) and one for Epson printers (at 720 PPI). When shot from a distance of 25-50x the focal length, the target will have higher resolution than the capture system, and it will be very hard (especially at 50x FL distance) to resolve all the way to Nyquist (which is located at a central blur radius of 92 pixels diameter). There is some more info about my targets here.

As for P1 support ... they do seem quite knowledgeable, but they aren't quick!  I wouldn't like to try to get an urgent problem fixed using email, but I suppose it would be possible to call one of their offices and ask to be put through to technical support.

BTW ... I have reported the adjustment 'issue' to P1, so it will be interesting to see what they say.

Good, they'll probably pass the info on to their software engineers, who will then prioritize based of frequency of reports and potential impact on stability and such. I've also reported non-critical issues in the past, and one is not yet fixed, others have been fixed in recent minor version updates. Some issues are hard to solve, because they are dependent on the (continuously changing version of) video drivers and the interaction with hardware acceleration, and hardware behavior can differ a lot between machines, not to mention operating systems.

Cheers,
Bart

[Robert Ardill]

I've been shooting a lot of shiny silverware and glossy ceramic objects lately, and have managed with an exposure adjustment layer to selectively 'tame' highlight clipping. Some clipping is okay, since it is just the mirror image of the lightsource which should Whitebalance to 'white' anyway, but it usually is nice to have a bit of the underlying design/colors in ceramics shining through.

Yes, the blown highlights in that image were reflections off shiny metal, glass and ceramic.

There are also other aspects of image quality that may be different between the Raw converters. I'm also quite pleased with the new resampling quality of C1 (LR was already pretty decent, unlike Photoshop).

Did you find that LR resampling was better for both up-sampling and down-sampling than PS?

I'll see if I can produce a comparison between LR and C1 with very small amounts of rotation and or keystone correction. Very small rotations or skews usually cause a fading in and out of sharpness (like it's shown here ( figure 8 )) that can look quite bad, and orientation and distortion correction are typical operations that trigger that.

Very interesting ... I look forward to your comparisons

Cheers

Robert
Cheers,
Bart
[/quote]

[Robert Ardill]

They are probably not telling the whole story (patented stuff), or do not know how to explain in a few lines. C1 probably adapts to the AA-filter being present or not. That could be interpreted as more or less sharpening, but I'd say that's not really what happens. It's more likely some different weighting during the demosaicing, to avoid AA-artifacts on some cameras. The A7RII is probably not pushed as far as a camera with an AA-filter. That BTW will probably mean that deconvolution sharpening also works better, because there are fewer artifacts that will be amplified along with the genuinely blurred signal.

Good point.

Yes, I think that is likely. An adjusted demosaicing weighting for cameras with and without AA-filter seem a logical approach. Aliasing artifacts are usually larger versions of the spatial frequencies that caused them. So it makes sense to avoid, or at least not amplify, them. Aliases have the largest modulation near the Nyquist frequency, and micro-contrast of the real signal is dropping to a minimum, so near the physical limit of resolution they show there ugliness 'best'. It makes sense to take special care and not make matters worse.

Well that may well explain why the 1DSIII Imatest results seem better near Nyquist than do the A7RII images (compared to LR, that is) ... perhaps C1 is deliberately toning down the signal near Nyquist whereas LR is not.

Yes, I have a version for Canon printers (at 600 PPI) and one for Epson printers (at 720 PPI). When shot from a distance of 25-50x the focal length, the target will have higher resolution than the capture system, and it will be very hard (especially at 50x FL distance) to resolve all the way to Nyquist (which is located at a central blur radius of 92 pixels diameter). There is some more info about my targets here.

Many thanks Bart ... I'll give this a go and let you know the results.

[Robert Ardill]

Continuing on with my evaluation of C1 v LR, I haven't been able to get the C1 development of this image to match LR's (the shadows, not the highlights and midtones):

R03998---.ARW

The LR- developed image looks like this:



I used a linear curve in LR in this case, but using a standard (DNG Profiler) curve works just as well.

For comparison purposes, in case anyone would like to give the image a go in C1, here is the LR-developed TIFF

R03998---LR.tif

Cheers,

Robert

[Bart_van_der_Wolf]

Continuing on with my evaluation of C1 v LR, I haven't been able to get the C1 development of this image to match LR's (the shadows, not the highlights and midtones):
[...]
For comparison purposes, in case anyone would like to give the image a go in C1, here is the LR-developed TIFF

Hi Robert,

Indeed, C1 cannot bridge the difference with a simple tone adjustment, but it can get a much better result when using an adjustment layer for the shadows (my interpretation attached). Unfortunately, it is also not perfect, due to limitations in the masking accuracy.

C1 expects less extreme illumination differences, e.g. by using fill flash indoors, or as I would personally approach it, with a dual conversion and a good HDR tonemapping program. This particular image would also be better when using a better masking program in Photoshop on a dual exposure conversion.

Cheers,
Bart

[Robert Ardill]

Hi Robert,

Indeed, C1 cannot bridge the difference with a simple tone adjustment, but it can get a much better result when using an adjustment layer for the shadows (my interpretation attached). Unfortunately, it is also not perfect, due to limitations in the masking accuracy.

C1 expects less extreme illumination differences, e.g. by using fill flash indoors, or as I would personally approach it, with a dual conversion and a good HDR tonemapping program. This particular image would also be better when using a better masking program in Photoshop on a dual exposure conversion.

Thanks for looking at this Bart.  I'm still very new to C1 so I thought that there was a reasonable chance that a more expert user could do a better job than me.  LR is very good at pulling the shadows.  Of course there is the question of noise ... which I haven't really looked at yet.

Robert

[ErikKaffehr]

Hi Robert,

Lightroom does a lot of what I (or they?) would call content aware local contrast enhancement. I think you get this whenever you move the "Higlights" slider below -50 or the "Shadows" over 50. It is sort of HDR like mapping.

This was introduced in processing variant 2012. Ta that time I experimented with blending HDR toning with normally processed images, but I mostly dropped that technique with arrival Lightrooms's PV2102.

Best regards
Erik

Thanks for looking at this Bart.  I'm still very new to C1 so I thought that there was a reasonable chance that a more expert user could do a better job than me.  LR is very good at pulling the shadows.  Of course there is the question of noise ... which I haven't really looked at yet.

Robert

[Robert Ardill]

Hi Bart,

I've tried your resolution test and it's very elegant and easy to use! Great!

Here is a result with the A7RII and Zeiss Loxia 21mm lens, photographed at around 1m.

C1 0.477 cy/px  105 lp/mm       Horizontal   
LR 0.45   cy/px    99 lp/mm         

C1 0.445 cy/px   98 lp/mm     Vertical   
LR 0.445 cy/px   98 lp/mm         

C1 0.453 cy/px  100 lp/mm    45 degrees   
LR 0.468 cy/px  103 lp/mm         

C1 0.44 cy/px     97 lp/mm   135 degrees   
LR 0.44 cy/px     97 lp/mm      



(LR on left, C1 on right)

So the measured resolutions are pretty much the same, subject to operator error, that is.  However LR clearly has much more severe color artifacts than does C1.

Cheers

Robert

[Robert Ardill]

Lightroom does a lot of what I (or they?) would call content aware local contrast enhancement. I think you get this whenever you move the "Higlights" slider below -50 or the "Shadows" over 50. It is sort of HDR like mapping.

This was introduced in processing variant 2012. Ta that time I experimented with blending HDR toning with normally processed images, but I mostly dropped that technique with arrival Lightrooms's PV2102.

Hi Eric,

I agree ... LR does a really good job with its Shadows/Highlights adjustments.  It's amazing how much the shadows can be pumped up.

Cheers

Robert