My ever-restless and creative friend Reena was recently talking to me about a series of photographs she’d taken over her last trip to a national park in her neck of the woods. They were just not turning out as dramatic and eerie as she wanted them to. This, of course, set my photographer-sense tingling and I wanted to know what the original pictures were like and why they weren’t working out.
She had already attempted a black & white version of one of the images of a rock pool, and while it was fairly dramatic, the flatness of the tones and a very bright highlight told me that the image was very likely overexposed. Since my curiosity had been piqued, I asked her to send me the original image to play around with. Since I was playing anyway, I thought I’d record my thought process as I tried to fix this overexposed photograph.
Exposure Correction Tools
The best cure for overexposure is prevention. Making sure you get a properly exposed shot from your camera is really the only true answer, and you can usually do that by setting the Exposure Compensation to a negative(-) setting in your camera (if you’re using one the automatic or semi-automatic modes). In daylight, at least, I find most digital cameras tend to overexpose, and a bit of negative exposure compensation does nicely.
Even with these precautions, however, the odd overexposed shot does slip through, so it’s helpful to know how to salvage them on the computer. For these corrections, a proper photo-editing application is the best way to go. I am well aware that there are now a slew of software tools like Aperture and Lightroom and the like, which make it much easier to handle large volumes of images with various automated pre-sets and complex features. They take a lot of the decision-making out of the hands of amateur who just wants a stunning image with the least amount of fuss. I don’t know anything about those and am more of the hands-on type of photo editor, so my choice of tool is a more nuts-and-bolts image editor which I use for all my professional work, and that is the GIMP (GNU Image Manipulation Program).
A layer-based photo editor gives you the best fine-grained control over the final image if you know what you’re doing, and if you’re serious about photography in the digital age, you would do well to learn your way around one of these. It’s very much like printing your own pictures in the dark room in the old days as compared to handing it over to a photo shop, even a professional one.
GIMP comes with a fairly complex collection of built-in features and capabilities but I will also be using one additional plugin script for the software called Shadows & Highlights, which well help us deal with some of the specific issues faced in badly exposed images. So if you want to try these methods on your own photos, I suggest you get a hold of these tools, which are both free, keep the GIMP manual open to figure out the specifics if you get lost, and follow along.
What is Overexposure?
The concept of exposure is at the very basis of photography. Photographs are formed or captured by light falling on a sensitive surface for a specific fraction of time. In the old days that surface was film, and today it is often a digital sensor. Either way, the concept remains the same; A certain amount of light energy is required to be captured to form a clear image. Dark scenes thus need a longer exposure, while bright subjects require a shorter exposure time. A dark scene that isn’t exposed for long enough turns out dark and underexposed, while a bright scene exposed for too long creates an image which is overly bright and overexposed.
This is exactly what has happened in this original image. The sensors in cameras that decide on exposure and other settings in automated modes, look at the average light quality of the scene in front of them and mathematically calculate the optimum exposure time. If a scene is very evenly lit, these averages turn out good results, but in scenes with drastic contrasts of light and dark, averages are often skewed away from the ideal. This is how, in an image like the one above, the camera can select a longer exposure to properly capture the dark and shadowy cave and rocks, which leads to the sunny parts of the scene being washed out with an excess of light.
This is apparent when you’ve looked at enough photographs and know how it works, but it is still useful to analyse these images more thoroughly to understand exactly what has gone wrong. One method to do this is to use a visual tool called an image histogram. This is a graph representing the amount of light, dark and medium value shades in an image. Many cameras can show you this read-out live on the scene, but once you have opened your photo in GIMP, you can display its histogram by going to the image menu Color > Info > Histogram. That displays the Histogram dialogue shown above.
The histogram represents the occurrance of shades in the image ranging from black on the extreme left to white on the extreme right. So as you can see from the histogram of our overexposed photo, it has no pure black or very dark shades at all, and you can see a sliver of the graph that shoots up on the white end, showing that some of the highlight areas of the image have been washed out into pure white.
This washing out of highlights is the more serious issue with overexposed images. Dark shades can be darkened, but once a highlight has been washed out into pure white, a phenomenon called clipping, there is very little that can be done to restore the visual information lost in those places.
The fact is, nothing in the real world is ever a mathematically pure white. There is always a tinge of colour and variations of shade in even the whitest of objects, so when something is captured as pure white in a photo, it is merely a symptom of not enough information being captured for that area. Look at the excessively darkened section of our image above and notice how the highlight foliage is beginning to turn red. That is because the pure white has no correct colour information, so the computer mathematically jumps to red when extreme darkening is attempted. There is almost nothing to be done about this particular problem, but first, let us tackle the broader overexposure problem.
Level Adjustment
The best place to start correcting exposure problems like this, is to remove that gap in the image histogram we saw earlier, and stretch the photo’s values across the entire range of dark to light. We can do this by bringing up the Adjust Color Levels tool through the image menu Colors > Levels.
The levels dialogue includes a histogram display, but this time with adjustments for the black point, white point and mid-point or gamma of the image, which are represented respectively by the black, white and grey triangles below the graph display. You can adjust the image by sliding around those triangles to new positions along the graph, or changing the numerical values in the boxes below them (0-255). For this image, the idea was to close the blank gap on the dark end of the graph so the black point was moved towards the point where the image data actually began. To solve the excess brightness of the image the gamma point was also moved to reduce the level of the middle values to something more natural looking.
As you can see from the resulting image above, this one adjustment to the photograph already gives it much more depth, and makes it look more true-to-life than the washed-out original. Levels adjustment is one of those essential tools in a photographer’s toolbox to improve almost any photo and bring it close to what was intended. Learn to use it well, because for a universally applied adjustment it pays dividends for very little effort.
Selective Layer-Based Image Improvements
Universal adjustments to image quality are all you need if you have a perfectly exposed shot, but as the scenarios get more complex and the subjects more varied, you need to start delving into layer-based adjustments to the whole or parts of the image. Some areas of a photo often need to be processed differently to get a good result and that’s where editable layers and layer blending come in.
In photo manipulation software, layers are like transparent plastic sheets stacked on top of each other. Things can be put on these sheets to super-impose them on whatever is below without erasing it, with what is below showing through the transparent parts.
Photo editing layers go further in allowing the top layer to affect the look of the layers below it by blending with them using various mathematical methods. There are a growing number of blending modes in most software, but there a few basic ones that are universally helpful. In Normal mode the layer simply super-imposes itself on the layer below, depending on the transparency of the imagery on the top layer, and also optionally on the opacity setting of the layer. In normal mode, layers work exactly like plastic sheets.
The three modes shown in the image above are particularly useful in photo editing. The three sections of the image were created by duplicating the image layer, so that there are two stacked layers with the original photograph on it, and then changing the blending mode on the top layer.
The Screen blending mode lightens the result depending on the lightness of the the image on the layer. A light area lightens more than a dark area. The Multiply mode has an opposite effect, darkening the final image, with the dark areas having a stronger effect than the light ones. The Overlay mode mixes the previous two modes, with light areas lightening and dark areas darkening the final image result. By applying these blend modes on duplicated layers of photographs, many exposure issues can be corrected with a great degree of control, by varying the opacity of the affecting layer.
Since layers can have only sections of the image and transparent non-affecting parts, it allows us to selectively apply these effects on only some parts of the image to make more localised adjustments. This is exactly what the Shadows & Highlights script mentioned earlier helps you with. If you downloaded and installed it correctly, it should be accessible through the Image menu Filters > Light and Shadow > Shadows & Highlights.
The Shadows & Highlights dialogue that appears has two settings which you can change but you don’t need to, because they have no permanent effect. The plugin works by extracting a blurred version of the shadow areas and the highlights areas of the image and creating new layers with them.
The new layers are set to the Overlay blending mode with the ‘fix highlights’ layer rendered in black and the ‘fix shadows’ layers rendered in white. The image above shows what the two layers look like if set to the normal blending mode. If you remember the earlier description of the Overlay mode, this means the ‘fix highlight’ layer darkens the highlights and the ‘fix shadow layer’ lightens the shadows. How much of an effect they have can be adjusted by changing the opacity of the layers and it is simply their initial opacity which is affected by the plugin settings.
For this image I chose to switch off the ‘fix shadow’ layer (click on the eye icon in the layers list), because the problem wasn’t with the shadows. The ‘fix highlights’ layer was kept on full opacity to bring back as much detail as possible from the light areas.
Trying to Recover Detail in Clipped Highlights
The clipped highlights are always the trickiest part to handle, because there is no amount of mathematics that can be applied to it to get back what was lost. Baring painting in all the missing details manually, there really is no way to recreate the missing detail, like in that one blown out branch in our overexposed image. That doesn’t mean the loss can’t be minimised, however, so I experimented with making the best of what was available. This entire section is entirely optional and may or may not even work for many images.
Since detail had gone missing, I thought I’d try to add in some random noise at least, to not have that part of the image be the only blank colour. I created a new layer filled with a middle grey, and then applied the RGB noise filter accessed from the menu Filters > Noise > RGB Noise.
The noise needed to only be applied to that one clipped highlight, so I right-clicked on the noise layer and created a layer mask using the Add Layer Mask option, setting it to white when asked. A layer mask decides which parts of a layer are opaque and transparent, without losing any actual image data. White makes the entire layer visible. The ‘fix highlights’ layer already had a selective representation of the highlights, so selecting that layer, it was copied (Edit > Copy), and then selecting the layer mask of the noise layer, the copied image was pasted (Edit > Paste). With the new floating layer anchored into the layer mask by right clicking on it and selecting Anchor. Now all the highlights were black on a white background in the layer mask, but since that would make the highlight areas of the noise layer disappear, the layer mask was inverted by selecting Colors > Invert from the menu.
All the highlight areas of the image now had the RGB noise appearing in them. Since I wanted them only for the branch in the top left corner, I selected the layer mask and right clicked on it to make it visible using the Show Layer Mask option. Then used the Magic Wand tool in the toolbox (with a threshold setting of 20 in this particular case) to select that white shape of the branch. I then used the Selection > Grow menu to increase the selection by a few pixels in every direction to make sure the highlight shape was fully covered.
To make sure everything else on that layer was transparent, I inverted the selection on the Layer Mask and then used Selection > Clear to make everything except the branch black. Now the RGB noise layer would only affect the branch in the corner.
Now, since pure balck & white noise isn’t exactly a natural occurrence, it was time to bring some colour into it. Making sure the RGB noise layer was selected and not the layer mask (notice the white border around the grey noise in the Layers list, rather than around the mostly black layer mask to its right), the Colorize dialogue was brought up from the menu Colors > Colorize. The Hue and Saturation slider were adjusted to create a leaf green. Here the saturation slider has been set to 100 to heighten the effect, but that should be lower to fit in with the natural colours in the image.
Perfectly sharp noise is also unnatural so the noise layer was then blurred using the Filters > Blur > Gaussian Blur dialogue set to 10 pixels to create a more muted noise effect. To complete this attempt to add detail to the clipped highlight, the noise layer was set to the Dissolve mode which creates a mosaic-like noise effect on the image below, and layer was also duplicated with the top copy set to the Burn mode to darken the highlight area using the noise as a template, with a lowered opacity to make it blend in.
The dual noise layers did a decent job of adding some detail to the highlight and even helped recover some of the edge details of the branch. I wouldn’t say the dissolve and burn layer modes are the definitive method to achieve this effect. In fact, the dissolve mode can be quite distracting in most images, but I hope this gives you an idea as to the sort of thinking that can help you experiment with the various modes to achieve the effect you need.
The final layer-adjusted image above is quite strong and would be enough of an improvement in most cases, but I thought the image could still be pushed a little more for that extra dramatic flair.
Monochrome Layers For Dramatic Images
No matter what political and social arguments can be put forward about the dangers of the supposedly new issue of manipulated images, there is no such thing as a real photograph. Every photo is a representation, either a limited one, or a heightened one, based on what the equipment is able to record. So there is no less validity in processing an image to create the desired effect as there is in trying to make it “real”.
Since this little experiment started with the need to create a dramatic image, I thought I’d try bringing in the power of B&W into the mix. While a conversion of a colour image into monochrome isn’t always desirable, black & white images do have a sense of drama and a graphic starkness to them that few will deny. So even if you don’t want to lose colour information, a monochrome layer can often help you push an image into pleasingly dramatic territory. I flattened all the layers that I’d created thus far and made a new image file to work on from this step forward. Such consolidation is often necessary when working on large images, for both your sanity and that of your computer hardware.
I duplicated the image layer to create a B&W version to play with. The simplest way to do the conversion is through the Colors > Desaturate menu, which will convert the current layer to monochrome using the setting provided. For many images, it can do a decent job of it, but there are other methods too.
The B&W conversion method that can provide the most dramatic flair, I find, is a GEGL operation in GIMP simply called c2g (colour to greyscale). It can be accessed in the menu under Tools > GEGL Operation, and then selecting c2g from the drop-down list.
This c2g operation can be very slow and processor intensive, so use it and make settings changes with care, especially if the live preview is enabled. It allows for 3 settings. Radius is a pixel size that is considered to create the range of greys. Setting it to around the width of your image will give you good results. Samples should be used with care because a higher number makes for a smoother image, but anything more than an increase of a few points can bring older computers to their knees. Increasing Iterations is also meant to improve quality, but the effect isn’t clear enough to bother with in most cases. Using mostly modest settings, I created a c2g version of the image on a new layer.
Considering the very high contrast look of the images produced by the c2g method, I set the layer opacity to 50% to temper the effect on the colour image, and blended the layer using the Darken only mode. This way the white areas of the c2g image wouldn’t effect the colours while the dark areas would deepen the shadows and provide more visual depth.
A final touch was to add a darkened vignette effect to the edges of the image. This was achieved by creating a new layer filled with a plain black. I then created a layer mask for it to control how much of it shows through. To create the vignette itself, the layer mask was filled with a radial black-to-white gradient with the whites towards the edges, so that the black fill of the layer would show through. The Colors > Brightness/Contrast adjustments were used to tweak the gradient as needed, and the vignette layer was then set to the Soft light blending mode to allow it to darken the edges without completely drowning them in black.
I am generally happy with the final result, because it corrects the problems with the image and also manages to achieve the drama that was missing. Having said that, it’s not perfect and I’m sure more tinkering would yield better results at various steps in the process described. Also, there are always options such as cropping to solve some of the issues, which were not considered for this particular walk though. The noisy highlight at the top left is an improvement, but the image would be a lot better if that top horizontal band of the photograph were removed completely to create a more wide-screen format image, for example.
These and many other improvements are possible and entirely different ones will be possible with the particular image you try these methods on. The purpose of this exercise has been less to tutor you in exact techniques and more to expose you to the thought process of trying to correct images. Once you understand more of that you will be better equipped to adapt your techniques to the need of the individual images and your intended effects for them.
Most of all, experiment, and learn your way around the software and the way layered editing works. Your photographs will be better for it.
Samir
