How to adjust for both clipped blacks and whites in filmic rgb ?

I (possibly) have an elementary understanding of how to adjust the parameters of the filmic module when my raw file contains clipped highlights, for example making use of the threshold control in the reconstruct tab to derive a reasonable image adjustment. But I don’t understand how to handle clipped blacks in the filmic module and would appreciate some advice.

Recent weather conditions in southern UK, with intensely bright (for UK) days and deep shadows in the hilly area where I live, have resulted in many of my photos having both clipped blacks and whites – sometimes in significant amounts. When I look at the histogram in filmic, I find that many of my images have an extremely long tail to the left in the blue channel. It seems to me that even if I have only 1 or 2 pixels clipped on the left, then the highlight reconstruct mask appears completely blue when I attempt to display it, making it unusable and preventing me from setting the threshold control optimally. If I attempt to eliminate the clipping on the left by use of the exposure module, then, because of the very long ‘tail’ on the left I mentioned, I have to greatly increase the exposure, dramatically increasing the clipping in the highlights.

There’s clearly something I’m not understanding about how to handle images that are clipped in both the black and the whites. I would normally expect to get suggestions about using PlayRaw, but then I receive ‘fixed’ versions of my images, the history stacks of which allow me to see what was done to ‘fix’ them. But this does not tell me the principles which lead to that set of adjustments being chosen – I have a ‘prescription’ for fixing those images, but haven’t learned enough to understand how to fix different images.

Any advice?

What do you want to do with the shadows? You can make them darker and loose that detail or try and recover some of that detail?

In broad strokes, move the Black Relative slider left to recover detail and right to darken the shadows. Stop moving the slider when you find the results pleasing.

I think he also has to make some photographic decisions about exposing the original shot to preserve what is the most important or to shoot some brackets…

By default your camera’s exposure will match to the mid-gray (18.4%) value. you can always recover shadow detail so that is (generally) less important but the highlights, critically, need to be kept within bounds.
My approach is as follows:

  1. First, compare a RAW with a JPG file. Generally the RAW data is 1/2 stop darker than the JPG that you see in your viewfinder. Preset the dt exposure to reflect this difference.
  2. Let your camera find its natural (18.4%) exposure.
  3. Use the camera’s exposure-compensation to avoid highlight clipping or to make adjustments for high-key or low-key shots … do not use the camera exposure setting to do this.
  4. check the dt exposure box for the exposure-compensation.
  5. If you start the filmic operation using ‘auto-tune’ and then adjust the 2 sliders to suit your taste.
    Working this way with filmic-rgb I rarely (almost never) need to touch exposure during development … filmic simply does it all.
1 Like

± some vendor-specific safety margin for highlights.

1 Like

If your images are really clipped, that is over-exposed to the point that pixels saturate, there is no fix, other than adjust your exposure in camera and try again. Same when you underexpose to the point of detail loss. And if you have a scene where you can’t avoid clipping at one end, you’ll have to use bracketing. (1)
That said, you can decide to allow some areas to clip (e.g. deep shadows or light bulbs), but then it’s a conscious choice and shouldn’t cause problems in post.

As for the long tail to the left of one colour: if you have a fairly intense green in your image (e.g. spring grass), there will be hardly any blue in that zone. The catch is in the “hardly”, combined with the logarithmic nature of the EV scale and the presence of noise: a small variation in energy can give a huge variation in EV in that range. “Clipping” such values in the histogram will not have a negative effect on the image. Otoh, pushing the black point to the left to avoid that clipping will get you a pale, washed out, image.

(1) another option could be to shoot earlier or later in the day, when contrasts tend to be less intense.

Yes, I had understood the option to bracket exposure and have been avoiding it up to now because I could not understand how to handle 3 or 5 images together in dt in an analogous way to processing bracketed exposures in, say, Photomatix. For example, choosing how to handle ghosts. But there are much bigger issues; algorithms like Photomatix seem to play about enormously with tones and luminance values and I have no idea how dt would react to those changes, or even if it would ‘see’ them at all, in a tiff produced by HDR software. Comments?

In fact I would like to see some wider guidance on how to handle bracketing (of any form) in dt.

Yes, I have learned about the ‘washed-out’ effect. And , yes, I appreciate that the number of pixels in the long blue tail to my images is in fact an insignificant proportion of all the pixels in my image - but they still have the effect of making the reconstruct mask a completely opaque intense blue colour if they exist at all, whether or not they have any material effect on the appearance of the image. How do I overcome that effect on the reconstruct mask ?

If that blue is when you have the over/under exposure indicator on, I don’t care about that, and I just ignore it. That blue indicates pure black in the mask (not in the image), and shows a region that will not participate in the reconstruction. Fine, it tends to be well away from the area to be reconstructed…

I use HDR Merge, which produces a DNG file. I then bring that DNG into darktable and process it normally.