AuntiPode wrote:
I like the processing ... with the exception of a few areas where there may have been a selection problem. Might also consider a USM 16, 60, 0 layer to bump the mid tone contrast.
Thanks Karen,
Interestingly, at least for me, the conversion and all of the global work was done in ACR, minor curves, and then use of the history brush in combination with varying blend/opacity settings. No selections were made - more care and practice in brushing technique needed.
Many functions are not available to use while in 32-bit mode. As such, it probably needs to be considered as a prep stage. To me, the gradation/refinement of tonal values seems finer (although that could just be "in my head") which could be helpful with dust bunny healing, and the ability to adjust gamma seems to have a bit more control over it, especially on a layer with additional opacity / blending refinement abilities.
It is also restricted to RGB (vs. CMYK or LAB).
Not sure just yet how much diff it is from doing the same on 16-bit. I think it makes it a bit easier to "see" what is in your file at a lower gamma, or set a lower 'base gamma' going into 16-bit ... then apply additional changes selectively rather than from a global gamma perspective. But, like I said, I haven't compared it with 16 bit, as this was really my first time playing with it ... and I certainly don't have any expertise here.
That having been said ... I'll likely explore it a bit more to see how it does / doesn't work into a workflow. I figure, it's there for a reason ... now just to understand how/why/when it has merit/value. Of course, so much more to learn ... no end in sight.
Realism in a B&W shot starts with a full range of tone, but that's difficult to capture in ambient light with a digital sensor due to its short range. That's a shortcoming evident in this shot, through no fault of yours. In most ambient lighting conditions outdoors we step to the plate with one strike against us. In the digital realm the best you can do is expose for accurate highlight rendering and add reflector or flash to try to adjust the scene range to fit the sensor.
By way of analogy if I was seeing the same results in print shot on B&W film my evaluation would be that it was underexposed in the shadows (which have lost detail) and underdeveloped relative to the range of the print paper used (which results in gray highlights). With digital, since exposure is pegged to the highlights on the fixed range sensor the evaluation, made "by the numbers" and visually using Levels in CS5 is underexposed highlights.
I held down the opt key (alt on PC) on the highlight slider and nothing clips in your shot, not even the specular highlights. Keeping opt pressed I moved the right-side highlight slider left forcing the lightest tones in the file to clip. At 208 the catchlights in the eyes and a few other highlights start to clip. The 256 point scale in levels relates to camera sensor as follows. Assuming the camera can record a range of 8 stops of detail, which is on the generous side, 32 units in Levels = 1 f/stops of exposure.
So based on when the limit clips in Levels my evaluation is that it was about one f/stop underexposed at capture. That is a less than optimal utilization of the already short DR of the camera. Not so much on the highlight side, because Levels can easily fix underexposed highlights, but in the shadows where every stop of range "given away" in the highlights by poor exposure control is a stop worth of shadow detail you'll never get back. That's strike two and that one is on you...
After correcting the highlight with the right-hand slider the midtones looked darker than I'd expect them to based on the other clues in the lighting so I moved the middle slider left to lighten them also. The left side if the histogram reflects the underexposed shadows. There's no way to fix it with the input sliders so I created a dupe layer from the original and adjusted the shadow output slider to the right, blended it into the first levels adjusted copy in the shirt and hat with a mask, winding up with this before / after result:
I used Levels here because I was starting with your posted JPG. Similar adjusts are better made at the RAW stage to try to "normalize" the tonal range, to the extent possible, before doing the conversion and other manipulation. At best it's getting to first with a walk vs. capturing a full tonal range in the camera, but that's the best we can do without flash assistance when scene range exceeds sensor range. However in this case if had the image been exposed more at capture per the clipping warning (just at the point the specular catchlights clip but skin doesn't) the camera most likely would have recorded the full range accurately as seen by eye.
As for 16 vs 32? Unless you are experiencing posterization and banding in tones or have a subject with very subtle tonal gradients when editing in 16-bit I doubt you'd see any difference in the output. Since the camera captures in 14-bit when you see banding problems with 16-bit editing it is usually the result of making gross adjustments to files in large editing spaces like ProPhoto RGB to overcome shortcomings in the exposure (i.e. lack of full tonal range) at capture. So in that regard 32 might be better in some cases, but if you start with a full-range capture you don't need to manipulate as much and don't create the problems the additional bit depth would fit.
Also, using a smaller editing space can reduce banding because there is less physical spacing (in terms of absolute Lab coordinates) between the colors. For example banding is a problem if using ProPhoto with 8-bit files. Switching to a smaller working space helps eliminate the banding. For a B&W image I don't see a compelling need for using Pro Photo.
The simple way is to find out if 32 bit is worth the processing overhead is empirically: try 16 and 32 on the same image with different working spaces and compare the end results.
This is where I think 32-bit might be the most helpful (hypothesis only).
Not really. If the camera does not record any detail in a shadow area due to its DR limit and the exposure more bits (i.e. creating more discrete tonal values) will not miraculously add detail.
Cameras sensors are like an array of buckets, filled in the highlights with a fire hose (lots of light) and in the shadows with an eye dropper (very little light). What signals the end of exposure is when the brightest light fill a bucket to capacity. The larger the "buckets" (sensor sites) are in capacity to store photons the long it takes to clip the highlights, which gives the sites in the shadows more time to accumulate light. The photons accumulate as analog voltage. When the exposure ends the sensor is read out site-by-site and the analog voltages are read and converted to digital values. But there is always some residual charge in the sensor sites —sludge in the bottom of the bucket — that varies from site-to-site, which manifests itself as noise in the underexposed shadows of the image.
What defines optimal exposure in the technical sense is a lighting ratio of key (dominant direction) and "neutral" fill which allows enough light to reach the sensor sites in the shadows to record detail by the time the brightest areas fully saturate and fill those sites — fitting the scene range to sensor.
That is easy to do with flash indoors:
1) Pick the aperture you want for DOF
2) Put a light just over the lens to create flat shadowless "fill" light and raise its power until you see detail in the darkest shadows of the scene.
3) Add an off axis "key" light to create modeling. Raise its power until a solid white object is just below clipping (around 250 eye dropper value).
What you will have done with steps 2 and 3 is fill the scene to the sensor perfectly and the resulting image will have a "seen by eye" tonal range in color or a B&W conversion.
It's not that easy outdoors in sunlight because the camera sensor can't handle the contrast in the shadows. In sunlight exposing for the highlights results in an image looking like this ..
But the flash overpowers the natural modeling and looks fake. Move the single flash off axis to get modeling and you have the same dark shadows in the areas the flash doesn't fit and the same scene not matching sensor dilemma.
The solution to matching scene-to-sensor outdoors is the same as indoors, at least in the flash-lit foreground: use two flashes in a key / fill arrangement...
You may not like to use flash outdoors for aesthetic reasons, or simply due to the logistics, but it or some similar supplemental light is needed whenever the ambient lighting contrast exceeds the capacity of sensor DR to record it naturally. The simple trick to making the flash look as natural as possible is to place the key flash so it hits at the same angle as the natural key light is hitting and modeling the object from above and to the side and keep the fill as shadowless as possible like the diffuse skylight fill we take for granted and don't even notice outdoors.
All the bit level does is define how many discrete steps there are in the gray scale between black and white. One bit 0/1 creates a two tone scale, two bits 00,01,10,11 a four tone scale, four bits 16 tones, eight bits 256... etc. At the 16 vs 32 bit level there are so many steps and the increments are so small the human eye can't resolve them all and see much difference. Both will look like a continuous tone gradient.
My hypothesis is that once you start applying algorithms for "correction" / "creation", even though the eye may not see the differences directly ... the penalty for algorithm upon algorithm upon algorithm may suffer a lesser penalty in the long run @ 32 bit than 16 bit (or at least some at 32 bit).
RustyBug wrote:
My hypothesis is that once you start applying algorithms for "correction" / "creation", even though the eye may not see the differences directly ... the penalty for algorithm upon algorithm upon algorithm may suffer a lesser penalty in the long run @ 32 bit than 16 bit (or at least some at 32 bit).
True on a technical level, but practically speaking if you can't see the difference in the final output it doesn't really matter.
Back just a few years ago when computer processors were slower and memory and disks were smaller and more expensive there was a huge performance penalty for just editing in 16-bit RAW vs. 8-bit JPG. Once computers became faster and memory and disks cheaper it became just as fast and cheap to process in 16-bit RAW and avoid problems like banding
The same is true with 32 vs.16 but I don't think the differences will be as evident in the final output, so practically speaking one is probably as good as the other if it doesn't slow down your editing process or cause you storage problems.
