ISO invariance is a comparison of an exposure-corrected image to a properly exposed image directly from the camera. A camera with an ISO Invariant sensor results in images that can be exposure corrected in post production while maintaining detail and noise levels similar to a shot exposed at the correct ISO from the camera. It’s called ISO invariance because ISO directly affects the amount of detail and noise in an image. The higher the ISO, the less detail and more noise. That’s the tradeoff to getting images when there isn’t a lot of light. Since ISO, aperture, and shutter speed control exposure, the ISO invariance of a sensor affects any sort of exposure correction even if it was not related to selecting the wrong ISO.
In the past, correcting exposure in post processing would quickly result in additional noise or loss of detail in the image. Had the image been properly exposed by using a higher ISO, it would have had more detail and less noise. Depending on the camera’s high ISO image quality, that may have been the difference between a usable image and trash. Sensors with higher dynamic range allow much more flexibility in post – so much that exposure corrected post images look very similar to correct exposure images direct from the camera. This provides many more recovery options for images that are underexposed.
Still cloudy? Let’s use an example. Say the correct exposure for an image is ½50 @ f/4.0 ISO 1600. I take an image at that setting. Then I fiddle with my camera and accidentally take a similar shot at ISO 100, underexposing the image by 4 stops. I bring both images into Lightroom, increase the exposure of the second image by 4 stops, and compare the two. On a camera that is nearly ISO Invariant (such as my Nikon D750) there will be very little difference between the images. They will have very similar levels of detail, and more importantly very similar levels of noise.
This would also work if I had used the correct ISO for the scene, but the wrong shutter speed or aperture. Say the correct exposure is ½50 @ f/4.0 ISO 100. I want to freeze droplets of water as a rock splashes into a pond so I increase the shutter speed to ¼000 but forget to adjust the ISO. My image is 4 stops underexposed but I can save it by correcting it in post. The result certainly has more noise than an image at ISO 100 but it barely has any more noise than if I had shot it at the correct ISO 1600 on the camera. Bottom line: I don’t lose the shot. The image at the top of this post is a great example. The original image is so dark that it is useless. By increasing the exposure by 3 stops and adding default noise reduction I get an image that is usable.
On a camera that is not ISO invariant (all Canons except the new D80), the post-processed image will contain visibly less detail and significantly more noise than the properly exposed image. I’d either have to live with the noise, soften the image by increasing noise reduction, or accept a darker image by reducing the exposure push by a couple of stops. The result is an image that is less than satisfactory. If you want a visual, DPReview has a fantastic example that demonstrates what I’m talking about. Compare the image at ISO 6400 (correct exposure) to the image at ISO 100 + 6 stops. They look pretty similar. Switch the camera to the Canon 5Ds R (released June 2015) and see the difference. The ISO 100 + 6 stops image has a significant amount of additional noise compared to the ISO 1600 image.
Image sensors have continued to improve their quality at high ISOs, allowing photographers to get great images in very low light, but exposure still isn’t perfect in every shot. ISO invariance allows us to basically fix the past without a loss in quality. My goal is always to get correctly exposed images, but it doesn’t always happen. Extra latitude in post is very important to me. It’s one of the reasons I decided to buy into a whole new camera system.