Adobe’s New Reflection Removal Tool Aims to Save Your Travel Photos

Removing reflections from photos takes time and effort. Adobe’s talented engineers are working to solve the issue once and for all in a single click.
Polarizing filters are an effective way to reduce or entirely remove reflections from objects like glass windows. However, not every photographer wants to carry a filter around with them all the time, so what can be done after the image is already captured? Historically, not much. Removing window reflections manually in image editing tools is tedious and woefully ineffective. However, armed with artificial intelligence, Adobe Camera Raw’s latest update solves this issue, rescuing countless shots.
AI remains a contentious topic among photographers for very good reasons. However, AI has the potential to save photographers significant time without wresting away creative control. Some recent features in Adobe Photoshop and Lightroom have maintained a nice balance here, including AI-powered masking, lens blur, and Generative Expand. Not every tool has hit the mark, but the new Reflection Removal one offers significant promise.
“Reflections are sometimes useful. For example, they help you distinguish between different materials. However, reflections that show recognizable objects are typically unwanted and get in the way of photography,” Adobe explains in a new blog post. “Indeed, we often give up taking a unique photo that we expect will be ruined by a reflection. So while few of the photos in our collections might contain unwanted reflections, this is mainly because we passed up many creative opportunities.”
Common examples of times reflections have a ruinous impact on photography include photographing landscapes through airplane windows, capturing travel photos at museums (that allow photography, of course), and maybe even capturing photos of animals at a zoo or wildlife rehab center. Windows and glass objects are everywhere, so there are countless times a photographer must contend with reflections.
However, sometimes reflections are themselves part of an image’s appeal. Few want to remove reflections of the sky from a lake, for example. Removing tiny reflections is rarely tricky to do manually. So, Reflection Removal is explicitly designed to deal with unintended, unwanted reflections that dramatically impact the look of the image. At least for now, more minor reflections are in the team’s sights for future iterations.
It’s a complicated problem to solve, though. As Adobe explains, a photograph “polluted by a reflection” is really the combination of two images. There is the scene the shooter wanted to capture, and then there is the second image, a reflected view of the scene behind the camera. The trick to removing a reflection from a photo is to disentangle these two different photos in a single frame. Adding noise to the issue is that reflections are not perfect. Glass is often dirty or smudged, interacts differently with light than the scene behind the glass, and usually has different degrees of focus and varying white balance.
“Also, in photos that contain reflections, if two edges cross, like the edges of the turntable and the legs of the photographer, it’s probably due to a superimposed reflection, because the edges of opaque objects don’t cross in nature,” Adobe adds.
Adobe trained its new Reflection Removal system using thousands of photos of different subjects without reflections. The team then combined pairs of these training images to simulate photographs with reflections. Since the team knew the ground truth of each picture, it could correct the model along the way, helping it improve over time.
“By repeating this training process over many examples, the model learns how to separate the two images in a photograph that is polluted by reflections,” writes Adobe.
As for the tool’s efficacy, it remains a work in progress. However, Adobe’s examples are impressive. Once the model finishes its work, the user can control how much a reflection is removed from the scene, from zero (no removal) to 100 (complete removal).
“Sometimes our model mixes up the scene behind the window and the reflected scene — that’s why we give you a two-ended slider,” Adobe says. “But because our model is not generative AI, it will never create objects that weren’t present in the original photograph.”
Although this new tool is impressive, Adobe says the “best way to remove reflections from a photograph is to avoid them in the first place.” But for times when that’s impossible, Reflection Removal shows promise.
For now, Reflection Removal only works with RAW photos, although JPEG, HEIC, and other non-RAW image support are planned. It is now available as a Technology Preview with Adobe Camera Raw, and the team anticipates bringing an expanded version to the entire Lightroom ecosystem in the future.
Reflection Removal is developed by Eric Kee and Adam Pikielny, with help from Jiawen Chen, Lars Jebe, Durga Ganesh Grandhi, Eric Chan, Thomas Knoll, Simon Chen, Frieder Gans, and Kevin Matzen.
“We believe our technology is the best-performing feature of its kind,” Adobe says. “So we’re excited to see what you do with it!”
Image credits: Adobe