A Full "Isoless" Implementation
A Full "Isoless" Implementation
Jun 8, 2012
I see the adoption of an ISOless AE mode as inevitable. ISOless mode is a logical extension of both TAv & M-with-AutoISO AE modes. ISO Sensitivity is not part of the sensor capturing process, because exposure is determined by Scene Luminance, Shutter Speed & Aperture. ISO is part of the image rendering process, along with tone response curve etc.
Using camera ISO requires commitment, as you reduce clipping headroom as you boost ISO. That may be OK with JPEG shooters, but for the raw shooter, using ISO where it is unnecessary (i.e when ADC referred input noise is low enough) and when EM hash pick-up in the noisy DSLR electromagnetic environment is well controlled, goes against the rationale of using raw: maximising flexibility & performance.
I see ISOess operation on a suitable camera as a superior form of highlight recovery, as whenever you are operating at an exposure combo where FWC is not reached for the vast majority of sensels, colloquially "operating above base ISO", there are highlights that have been captured and not yet discarded. (This discarding occurs when ISO is boosted and a higher percentage of the image highlights clip in the ADC.)
Here is what I think is required for a full ISOless implementation:
1. An extention to Camera AE modes, raw file formats and raw converter operation.
2. A normal-brightness in-camera review image - the review image on your camera's back LCD screen would be boosted an amount determined by the camera's metering to produce a "correct" gain in a similar manner to how metering works now, so you have a(This assumes you have not reached FWC i.e. be in a lighting situation where, for your choice of aperture and shutter speed, you would normally raise the ISO.) The embedded preview JPEG would similarly be recorded at this higher brightness rendered level.
3. The camera could also show this initial boost gain figure in stops in the viewfinder as you are composing. So you are aware how much boost your current aperture & shutter speed combo require to produce a good brightness level. Like showing the current ISO in an Auto-ISO mode. Obviously, the higher this figure, the more noisy the resultant image will be, in a similar manner to currently using high-ISO. You could even show an "ISO number" instead of stops of gain if that makes it more friendly to traditional photographers. So the camera display in ISOless mode could could show either "Stops Gain" or "ISO" depending on a camera menu option.
There is no no need to have discrete steps in this gain, but in the "ISO" display mode, standard part-stop steps could be shown.
4. When the raw file is written it also has this "initial" gain setting included as meta-data in the raw file. So when you open it in your raw converter you will see the image normally. Also, if you do a "default" conversion, it will use this figure. However you can freely adjust away from this suggested initial gain.
The ISOless operation in both camera & software would be largely transparent.
5. If a high-contract scene with the same exposure is shot both at say ISO3200, and at base sensitivity (say ISO100), the later will have more highlights captured than the first, but if we do a straight 5-stops gain boost in PP we'll end up clipping a similar % of highlights (assuming a similar TRC in both). To make best use of this extra headroom, the raw converter should offer various levels of aggressiveness of highlight compression and percentage of pixels clipped controls. Maybe logic in the converter should know when to use traditional highlight recovery (probably close to FWC) and when to use this better method.
Also it should probably have a gain adjustment range of 6-8 stops to cover the traditional ISO100-ISO25600 range to suit both APS-C & FF.