DR Correction

[Swa]

I'm wondering if someone can explain this to me. I've been playing around with dynamic range correction a bit. What I've noticed is that on the higher settings it also limits the ISO you can use. This would make sense if it was processing the image.

Some of the photos on the SX50 HS in bright light look really horrible. Like they are overexposed but at the same time still noisy. I can't see how increasing the ISO and so also the noise helps. Can the same effect be achieved processing from RAW or does the higher ISO also play a role?

 

[fvdbergh]

I'm wondering if someone can explain this to me. I've been playing around with dynamic range correction a bit. What I've noticed is that on the higher settings it also limits the ISO you can use. This would make sense if it was processing the image.

Some of the photos on the SX50 HS in bright light look really horrible. Like they are overexposed but at the same time still noisy. I can't see how increasing the ISO and so also the noise helps. Can the same effect be achieved processing from RAW or does the higher ISO also play a role?

I am not familiar with the SX50, but "dynamic range correction" sounds like the similarly named feature on DSLRs. Here is what those modes try to achieve:
1. If you just increase ISO, then you clip highlights (which are irreversibly overexposed). So you keep the ISO low.
2. Underexpose the image on purpose, thereby adding additional protection against overexposing (small) highlights.
3. Boost exposure in software (just like the "exposure" tab in ACR/LR) to compensate for the underexposure.

While this recipe does not sound all that good when explained like this, it actually works. You can always add a bit more noise reduction to compensate for the increased noise (especially in darker parts of your image), but you cannot recover areas that were hopelessly overexposed. Since your camera's auto-exposure metering is usually performed at a reduced resolution (separate sensor on DSLRs), it is actually quite common for the camera to miss small highlights, thus causing the camera to select an exposure that will overexpose said small highlights.

Using the "dynamic range correction" option allows you to capture such tricky high-dynamic range shots in-camera using a single shot, kind of like HDR-lite. This comes at the expense of image quality, so it should be used with care.

 

[Swa]

I don't understand #1. As I understand it ISO controls the sensitivity of the sensor so without also selecting to overexpose the image it won't be overexposed just by increasing the ISO. So the question still remains of what exactly increasing the ISO does. The camera doesn't limit you to lower ISO but to higher ISO.

As it stands I think there's some tinkering with the actual electronics involved here that can't be achieved in software alone after the image is taken. What I'm trying to figure out is what the ISO does as I'm not aware of any effect besides increasing noise. So I don't know if shooting at a higher ISO will have an effect when adjusting in software. As I stated the problem isn't just overexposure in these areas but also an increase in noise which makes it look really awful.

 

[fvdbergh]

I don't understand #1. As I understand it ISO controls the sensitivity of the sensor so without also selecting to overexpose the image it won't be overexposed just by increasing the ISO. So the question still remains of what exactly increasing the ISO does. The camera doesn't limit you to lower ISO but to higher ISO.

As it stands I think there's some tinkering with the actual electronics involved here that can't be achieved in software alone after the image is taken. What I'm trying to figure out is what the ISO does as I'm not aware of any effect besides increasing noise. So I don't know if shooting at a higher ISO will have an effect when adjusting in software. As I stated the problem isn't just overexposure in these areas but also an increase in noise which makes it look really awful.

At the hardware level, the signal coming from the photosite is an analogue one. Before this signal (voltage corresponding to number of photoelectrons captured by the photosite) is converted to a digital number by the ADC, the signal is amplified. The purpose of the amplification is to increase the signal level to reduce the relative magnitude of the quantization error, and possibly read noise (depending on the sensor design). The amplification of the analogue signal is called the gain, and typically (traditionally?) this gain factor is what is increased when selecting a higher ISO setting.

Things got a little more complicated after Sony started producing sensors where the ADC is integrated into the sensor itself, which resulted in the so-called ISO-less cameras, such as the Nikon D7000 (and many others, since). With these sensors there is little benefit to adjusting the analogue gain on the sensor --- you can just as well simulate a higher ISO by multiplying the digital number (after the ADC process) by a factor > 1.0. Strictly speaking, these sensors are only approximately ISO-less. On these sensors, the method I described in my first reply (above) makes perfect sense, since you effectively lose nothing by using a lower ISO (lower ISO followed by boosting in post, vs. higher ISO, in terms of visible noise). On the other hand, using a lower ISO effectively prevents clipping of highlights.

To get back to your original question: I have no idea what SX50's DR correction does, but it is most likely what Nikon calls "Active D-Lighting" on its DSLRs, and what Canon calls "Highlight Tone Priority" on its DSLRs. Although it seems strange that this option restricts you to a higher ISO, it is functionally consistent, i.e., if you are shooting at ISO 200 with DR correction enabled, the camera would choose to keep the shutter speed fixed, and drop the ISO to 100. If, say, ISO 100 is the lowest ISO setting of the sensor, then the camera must limit you to ISO settings of 200 or higher when you turn on DR correction, in order to give itself room to "drop the ISO".

Let us assume you set the camera to ISO 200 with a shutter speed of 1/100 s, with DR correction enabled. You could produce similar results by keeping the ISO fixed at 200, and doubling the shutter speed. This would produce exactly the same amount of underexposure (ISO 200 at 1/200 s produces an equivalent exposure to ISO 100 at 1/100 s), but doubling the shutter speed affects other aspects, such as the degree to which motion is frozen, hence the camera fixes the shutter speed, and drops the ISO "behind your back".

I would recommend reading up on the sources of noise for further insight. Understanding photon shot noise will make the operation of "DR correction" clear.

 

[Swa]

Ok that makes sense. So the camera doesn't really use the reported settings but uses different ones? So if it's set to ISO 200 it really uses ISO 100 to underexpose the image and then does the corrections without incurring the clipping?

Am I right then that it may be better to just reduce the exposure manually a bit when shooting in bright light?

 

[fvdbergh]

Ok that makes sense. So the camera doesn't really use the reported settings but uses different ones? So if it's set to ISO 200 it really uses ISO 100 to underexpose the image and then does the corrections without incurring the clipping?

Yes.

Am I right then that it may be better to just reduce the exposure manually a bit when shooting in bright light?

Probably. The thing to remember is that the camera tries to make an educated guess as to what the ideal exposure is, based on what the metering sensor measured (although it is possible that a non-DSLR camera could use the main image sensor to calculate exposure). The camera is likely to follow the ETTR (Expose To The Right) rule, i.e., try to maximise the amount of light gathered, subject to whatever mode you have selected (aperture priority, shutter priority, "program exposure", auto, whatever). This type of heuristic will occasionally clip some highlights; the DR correction mode is a way of fine-tuning the exposure calculation to hedge against clipping the highlights, at the potential risk of introducing more noise.

So it is entirely possible to select a better exposure than what the camera estimated. Strictly speaking, exposure, just like composition, choice of focal length, etc. is an entirely subjective "artistic" choice, so if the camera is not choosing the parameters that you want, you can always override it with manual controls.

Just to cover all the possible angles, keep in mind that selecting a faster shutter speed (or lower ISO) to always force underexposure in bright light is not going to be the correct choice for all scenes, i.e., use your own judgement.

 

[Swa]

Ok thanks. The problem is mainly in direct sunlight. So selecting an exposure a third stop below isn't really going to change things too dramatically and the images may even be better if it looks like they were taken in the shade. Perhaps I should just also try using spot exposure instead of evaluative.