SHOOTING HANDHELD IN LOW LIGHT
During the film era, it was impossible to take a photograph in low light, without a tripod or a flash. That was because the ISO (or its equivalent predecessor, the ASA) rating of of the films being used was just 100 and sometimes, even lower. Images from faster films were of low quality and in any case, the ISO maxed out at about 200, unless you did push processing, which made the quality even worse
Even after digital cameras were introduced, the low light capability was nothing great till larger sensors and CMOS technology brought a sea of changes. Now, it is quite common to see cameras offering ISO values in six digits and some even going over a million.
Whatever may be the case, you will sooner or later encounter a situation where you feel that the light is not enough to take a picture with your camera handheld, and you do not have a tripod or a flash (or even if you have the latter, it is of no use for example, if you are photographing a monument or a landscape). Under such circumstances, you would do the following. Crank up the ISO to the maximum level, switch on image stabilisation and set the lens at the widest aperture (smallest f/ number). If possible, you will try and use a shorter focal length to get a slower handholdable shutter speed to satisfy the standard thumb rule. That’s about it. Let us say you have done all that and you look at the shutter speed on your camera only to find that it is second or slower. Not enough to handhold
So, what next Pack your bag and leave No, not necessarily. Here is a technique that the author has been using with good results and this article describes just that.
This technique relies on a few important underlying facts.
Modern digital cameras, especially those with large sensors, have a huge dynamic range and consequently, a very large latitude. Recall that dynamic range is the brightest and the darkest tones that your camera can simultaneously record in a single frame. Latitude means how much wrong exposure can be given at the time of image capture and yet, we can extract a useful picture from that image. In digital cameras, modern sensors allow a huge amount of latitude for underexposure, up to 4 or 5 stops (Picture 1). Also, for your information, digital cameras have very little latitude for overexposure, but, this is not relevant to us in the current scenario.
That is, you can underexpose an image even up to 5 stops and then increase the exposure in post-processing (Picture 2) to extract an acceptable picture. However, there is no free lunch and when you do so, there is a large amount of noise that is introduced. Though this may not be objectionable on social media, it will be unacceptable for a good quality print.
The next point is the nature of noise in digital sensors. Let us look at this in more detail. By the simplest definition, noise is something that alters and thus corrupts what you want. Those who study electronics are told that any ‘unwanted’ signal contributes to ‘noise’ as it corrupts the needed signal.
As photographers, we want artefacts (things that are not originally in the scene that is being captured) to be absent in our captured image, as these degrade the overall quality. Noise in images does just that - it adds something that is not there originally. Noise has two components, called ‘Chroma’ and ‘Luminance’. The former causes pixels of uneven colour resulting in coloured blotches (mottling). The latter gives rise to uneven brightness of pixels that cause a grain-like appearance when you were expecting a smooth result.
Most popular post-processing packages like Photoshop or Lightroom have commands with very advanced algorithms to reduce both these types of noise. Plus, many specialised applications do noise reduction. All these have a problem. They do reduce the noise, but, in the process, they also reduce other details and thus make the image soft. Is there a way out
Yes, there is, and for that, we first need to remember the nature of noise in digital sensors. This is the random character
(non-uniformity) of noise - a technical fact that we are going to exploit. Try this experiment. Fix your camera on a solid tripod, set your camera to a fixed high ISO value, and take ten photographs of a static scene in the same light with the same exposure. Now, if you see each picture magnified very large you will find that the noise will be present in each picture but the pattern, that is the location of noise will differ from image to image. Thus, if we take all these images then somehow average them, then those parts which are static (that is, the objects in the scene) will remain unaffected but the noise will be reduced drastically as it will average out due to the randomness. Taking the average is also called calculating the Needless to say, more the images you have the better will be the result.
‘mean’.
Another advantage is that, when averaged out, noise is reduced but the detail is not lost, as we are not smoothening any area, we are only acting on the randomness. For more details, see the article ‘Noise Reduction Through Multiple Exposures’ Smart Photography, April 2018 issue.
So far so good. But how do we put all this into practise to reach our goal of handheld photography in low light For the sake of clarity, we will divide the whole process into two parts, i.e. image capture and post-processing. Read on for more details
Image capture
The method of using this technique is best illustrated with an example. Let us start with what we said at the beginning of this article. We have a low light scene and after we did what all we could, we need a shutter speed of 1 sec to get a good well-exposed image, but the slowest shutter speed you can handhold under the circumstances is, 1/ 0 sec (as an example). This is equivalent to 5 stops of underexposure. If you take a photograph at 1/ 0 sec, you will get an image that is grossly underexposed (Picture 1). If you now boost the shadows, you will get an image with proper tones (Picture 2), but it will have an unacceptable amount of noise.
What next Well, here are the steps to get over that noise issue.
Use raw file format. If this is not available, use the best quality PEG available.
Switch off the in-camera NR. This will only make the images soft.
Do not use the auto white balance. Set it to a known value (example, shade, cloudy, etc.) depending on the situation. Better still, do a custom white balance. In all these cases, you will get consistent colour across all the images
- something that cannot be guaranteed with auto white balance. So, if you have photographed in raw, the same colour correction (if needed) can be applied for all images.
Select either aperture priority or manual exposure mode. Choose a shutter speed (with the help of a suitable aperture and ISO) that you can safely handhold. In our case let us say it is 1/ 0 sec.
Keep the ISO and exposure fixed at that value for all the images in the sequence.
Select high speed (maximum that your camera permits) continuous frame mode (Picture ) so that the capture of images is completed in the minimum possible time, thus reducing any inter-frame or subject movement.
Recall that for our example, the correct shutter speed is 1 sec and what you can safely handle is 1/ 0, which is 5 stops under. So, you need to take at least 5 images, preferably even more in a burst.
How does this help We now have
(at least) 5 images, each of which is underexposed. If we increase the exposure in post-processing, then the images will be very noisy as mentioned. However, we have taken 5 images (together called a stack), so we are capturing 5 times the light cumulatively, as compared to a single capture. That is all The rest of the trick is in post-processing.
Post-processing
Open Lightroom. Add the stack to the catalogue by importing it through the Library module. Invoke the Develop module by pressing the ‘D’ key. Select all the images in the stack. At the bottom right of the Develop module screen, you will find the button Auto-Sync. In case you are seeing instead of this, a button labelled Sync, click on the small switch and the button will change to Auto-Sync.
Click on it.
Now open up the shadows, increase exposure, and anything else you wish to do to get the image to an acceptable level of brightness with the right tones and colours. Since Auto-Sync is active, any change you make will be applied exactly the same way to all the selected images. In other words, you are simultaneously editing all the images in the stack together, identically. Keep the edited images selected.
Once you are satisfied, from the main menu of Lightroom, choose,
Photo > Edit > Open as Layers in Photograph
(This is the last option in the dropdown menu)
This will launch Photoshop (if it is not already running) and each image file you have captured will appear as an individual layer in a single Photoshop file.
Before you proceed further, there is one very important step that you need to take. Since you are photographing handheld, there will be a slight movement of your hands and/or body between each frame. Consequently, your camera would have (slightly) shifted up or down or right or left. Also, you might have rotated the camera or moved it back and forth involuntarily, while the burst capture was in progress. However, for our technique to work, it is absolutely, repeat, absolutely necessary that the images are aligned. To do this, open the Layers panel in Photoshop by pressing function key (F7) (Picture 4).
From the Layers panel, select all layers by clicking on each layer while keeping CTRL key pressed. Alternatively, click on the first and the last layer while keeping the Shift key pressed.
Now go to the Photoshop main menu, select
Edit > Auto-Align Layers
You will see a screen (Picture 5).
Select Auto and then click on OK. Photoshop will now align the layers and in the process, may shift or rotate the images a little bit. So, the layers may look a little skewed and you may have to crop the overall image to get a proper rectangular. This is why you should frame the images during capture, a little liberally, so that you will not cut off anything important.
Once again select all the layers (as just described) and from the Photoshop, main menu, choose
Layer > Smart Objects > Convert to Smart objects
Layer > Smart Objects > Stack mode > Mean
Save the resulting file. It will be automatically added to your Lightroom catalogue. That’s all to it Be aware that both these steps take time, up to several minutes. Please study Picture 6 carefully. It shows three photographs side by side. The left section shows the photograph taken with the correct exposure and the one at the centre shows the noise that builds up when you take the underexposed image and boost the exposure. On the right, you can see the photograph of the processed image, which is the result of taking a mean of several images. You can see that it compares favourably with the photograph taken at the correct exposure. There might be a slight softening of the image, but it can be sharpened a little, as it is free from noise. All three photographs are shown at 100 here.
So, what has happened When Photoshop calculates mean (average) it eliminates noise (as it is random and hence the pattern is not identical across all frames, as mentioned), whereas, the main objects that are common to all the frames are retained. Thus, you can now get an image with noise reduced while retaining the detail.
Limitations
As you can see, this method is useful only if your subject is static like interiors, architecture, landscapes, portraits (if the subject is willing to remain still just or a second or two), etc. If any object in the scene has moved, the final processed image may exhibit some artefacts (called ‘ghosts’). How objectionable these can be, will depend on the total sum of the subject’s presence (time) in all the frames put together and the speed at which the object was moving. If the subject has moved very rapidly, then the registration may not be very much in each frame.
You can also remove the artefacts in post-processing before or after you process the stack, depending on their nature. I have even shot street scenes with this technique. Also, do not underexpose by more than 4 or 5 stops as this technique may need too many frames to average and remove noise.