[GUB]

So I checked the result using Pentaxs own camera utility 5. Chose the 2 stop overexposed image (1890) and set it at -2 EV to bring the histo out. Result is showing massive clipping similar to Darktable.
So once more I reiterate - would Pentax really have a couple of stops of dynamic range spare without utilising it in its own software?
I suggest that the highlight reconstruction being claimed is merely extrapolation of the existing tonal range (of lower valued pixels) utilizing the fully saturated pixels .

Attached Images

 

[GUB]

Does that mean that you are willing to attempt recovery?


Steve

In this example he may actually do a fair job of it Steve. With the weave of the cloth presenting lots of samples of unclipped tonal range beside the clipped areas the computer can probably do quite a good job of guessing the values in the clipped areas. Because a guess job it is.
colour / highlight reconstruction in Darktable;
Color Reconstruction | darktable

[TonyW]

Does that mean that you are willing to attempt recovery?
Steve

Have you examined the posted images? If so apart from normalising The rendering what exactly needs recovering

---------- Post added 05-22-19 at 12:36 AM ----------

That is the theoretical value. The assumption is that the chip supports the full 14 bits for all channels.


Steve

That is indeed a theoretical value and not all cameras will reach it may only get to 16000 or so. But unless the Pentax is a special case scenario once saturation point hit I have never seen a case yet where the data does not hit the wall and climb in one of several shapes. If this is the case for Pentax so be it and I stand corrected, my 645z does not exhibit the behaviour of a gentle tail off when saturated the wall is there.

I was under the impression that you actually intended to post an image with full well saturation?

[TonyW]

Just the one image that you claim overexposed (assume by this you mean from the metered reading?). From what you said about your method of placing red value to the right in camera histogram (you are increasing exposure) then subsequently you are increasing exposure by +0.3EV your value of +2EV is probably actually close to +3EV as shown below


Shows zero signs of clipping either in Rawdigger of ACR - I have never seen clipping without an associated histogram climbing the wall. See attachment of a view of your raw data for this image. You may also note that the red value is nearly at +3EV and you may also wish to check back what I said about taking a meter reading of a specific area and increasing the suggested exposure by +3EV which may get you very close to saturation - this represents a meter calibration point of 12.5% which may be representative of your system, there may be even a little more at a pinch

As already hinted the raw editor application of profile can make a huge difference to how the render appears and this will be reflected in the histogram. The attached gif once again showing no clipping even with one of the Adobe standard profiles or the Pentax embedded profile in place. Bearing in mind this not taking into account the fact of adjustment by the editor under the hood.

Now should the colour or any other aspect be wrong in your opinion then it should be obvious what you need to do to correct in your application. If not then you may indicate a ROI that must be a particular colour and supply the values for that area. I suggest Lab values as then there is no colour ambiguity when using different working spaces

There is no guessing here. And as you seem to want to be mildly agressive I will reiterate that your testing and conclusion appear to be flawed.

Including this nonsense which does not deserve an answer.

So once more I reiterate - would Pentax really have a couple of stops of dynamic range spare without utilising it in its own software?

Attached Images

   

---

Last edited by TonyW; 05-22-2019 at 02:54 AM.

[GUB]

Just the one image that you claim overexposed (assume by this you mean from the metered reading?). From what you said about your method of placing red value to the right in camera histogram (you are increasing exposure) then subsequently you are increasing exposure by +0.3EV your value of +2EV is probably actually close to +3EV as shown below


Shows zero signs of clipping either in Rawdigger of ACR - I have never seen clipping without an associated histogram climbing the wall. See attachment of a view of your raw data for this image. You may also note that the red value is nearly at +3EV and you may also wish to check back what I said about taking a meter reading of a specific area and increasing the suggested exposure by +3EV which may get you very close to saturation - this represents a meter calibration point of 12.5% which may be representative of your system, there may be even a little more at a pinch

As already hinted the raw editor application of profile can make a huge difference to how the render appears and this will be reflected in the histogram. The attached gif once again showing no clipping even with one of the Adobe standard profiles or the Pentax embedded profile in place. Bearing in mind this not taking into account the fact of adjustment by the editor under the hood.

Now should the colour or any other aspect be wrong in your opinion then it should be obvious what you need to do to correct in your application. If not then you may indicate a ROI that must be a particular colour and supply the values for that area. I suggest Lab values as then there is no colour ambiguity when using different working spaces

There is no guessing here. And as you seem to want to be mildly agressive I will reiterate that your testing and conclusion appear to be flawed.

Including this nonsense which does not deserve an answer.

I am a little bit confused - is one of those images in your gif supposed to be corrected?

[GUB]

GUB's red undies -- taken from image 1884 - the image I consider correctly exposed. Plus a 100% crop.

[GUB]

Image 1890 - 2 stops over what I consider ideal exposure and one stop clipped as far as I am concerned. Highlight reconstruction applied in Darktable.
I feel Highlight reconstruction has done a great job considering it is guessing the tonal ranges it applies to the clipped areas (which the literature freely acknowledges).
First time I have used it - may use it more in the future when I find I have stuffed up the exposures.

Attached Images

View Picture EXIF

File Date

 

View Picture EXIF

File Date

 

[TonyW]

I am a little bit confused ...

That is clear and you also seem to want to move the goalpost. My image not corrected but showing that you do not have clipping or full saturation in LR and confirmed in Rawdigger. Your posted 'corrected' images bear no resemblence to your original screenshot and in the process of editing you have lost the yellow/gold thread next to the dark blue, this may or may not be intentional!

This is not as much about image colour as approaching saturation/ clipping and yet you still choose to ignore the evidence in Rawdigger as it contradicts you theories. Presenting as proof your image file in DCPU is flawed as you are still not getting it - you are not looking at the raw data you are looking at a rendering of that data in DCPU and in this case the rendering chosen favours the look of clipping and yet I bet that if you actually edit the image that you will see that the red channel has not hit a wall

I did say that:

Now should the colour or any other aspect be wrong in your opinion then it should be obvious what you need to do to correct in your application. If not then you may indicate a ROI that must be a particular colour and supply the values for that area. I suggest Lab values as then there is no colour ambiguity when using different working spaces

This is purely an editing issue but with a full unclipped file should not present any problems

Below first is your orignal image in your editor with your newly edited version below. You have made a huge change to the reds, but that is your perogative!
Second image is a quick edit in ACR not wasting any time trying to match your edit in any way but showing again that there is no clipping.

[TonyW]

Just a quickie to show an image of a very red jumper capture 645z deliberately over exposed showing clipping in Rawdigger of the red channel only.

Looked worse on initial import into ACR due to Adobe applied adjustments but red channel still clipped. Fairly straightforward correction to pull back to look the same as the more conservative 'correct' exposure. The ACR colour represents the jumper colour fairly accurately on a wide gamut monitor and can be reproduced closely on some colour print materials

More severe clipping may be recoverable with some work in PS. Just to be absolutely clear I AM NOT advocating exposing to clipping point on the contrary get it right in camera has got to be the best advice.

[Ian Stuart Forsyth]

ere is the 7 test images for the test to establish clipping point.

I had a look at the image with the largest exposure in the group 1890 and it is still around one stop from clipping the red channel

[TonyW]

I had a look at the image with the largest exposure in the group 1890 and it is still around one stop from clipping the red channel

Thanks for the confirmation which now suggests the image chosen as a 'correct' exposure is actually two stops from clipping the red channel. So effectively losing 2 stops of DR !

[Ian Stuart Forsyth]

Thanks for the confirmation which now suggests the image chosen as a 'correct' exposure is actually two stops from clipping the red channel. So effectively losing 2 stops of DR !

I will add that there has been a shift in the temperature of the light in between the 1st and the 7th image as you can see in the raw histogram, it is slight but it is there .
This has changed how the areas that are in more direct light from above show a slightly different color cast, while the areas in that are mainly being supplied light from the reflected surfaces from around the sweeter don't have this color shift. this would tell me that there is a stronger light source for the last image from directly from above is warmer.



Here I have WB for that difference and just applied a tonal adjustment to reveal the detail in the red sweater ( just quick tweaks) , one side the first image and the other the 7th image mirrored. Other than motion blur and different focus the 7th is fully recoverable with a little more wiggle room allowed in the exposure before the raw is clipped.

---------- Post added 05-23-2019 at 03:02 AM ----------

Image 1890 - 2 stops over what I consider ideal exposure and one stop clipped as far as I am concerned. Highlight reconstruction applied in Darktable.
I feel Highlight reconstruction has done a great job considering it is guessing the tonal ranges it applies to the clipped areas (which the literature freely acknowledges).
First time I have used it - may use it more in the future when I find I have stuffed up the exposures.

Where you think you are showing chipped and what you think is clipped is not from being overexposed and blowing the red with a that exposure , where a lot of people think reds are overexposed are not found in the areas contained with in the raw files reds.

For this image I used a cc30m filter to pre WB the light so that the red channel is closer to the green channel as to where they would clip this way I can maximise the red channel information I can store in my sensor.


here is what it would look like without this prewb filter

my greens are way out in front to clip in normal daylight first.

When I have done everything to put the priority on clipping the reds it does not show up in the red that you have posted above, if you look at my first image I have clipping shown as blue patches in my first image.
That bump in my red channel in the first image is all of that data where there is the very lightest pinks around that blue.

For me this would be the perfect exposure for the raw file when shooting this flower.