MIT reseachers remove unwanted reflections from photos

Taking photographs through a window can be a mess. Anything behind the camera could be reflected, infiltrating the photograph. It can be difficult to see the intended image due to the photographer's own reflection obscuring the image. In the fictional, TV world of CSI, you could just say, "enhance!" But, until now, it hasn't been so easy to erase photographed reflections in the real world. Thankfully, a group of MIT researchers have figured out a complicated algorithm which can remove undesired reflections from photographs.

The algorithm can automatically detect and remove reflections from digital photos. It can even isolate and preserve the reflection, splitting the compound photo into two images: the reflection and the intended image.

The scientists found that the most troublesome reflections are created by double-paned windows and those with thick glass which create two sets of reflections. To identify areas where reflections intruded upon an image, the team identified small clusters of pixels that abruptly changed colors. The team then sought the boundaries of the color changes, isolating the reflection.

The group hopes to the new algorithm can be applied in everyday photography. In the future, it could be incorporated into photo-editing software toolboxes or integrated into a camera's onboard software, like current image stabilization features. Additionally, the formula could find its way into military applications, allowing drones and remote-controlled robots to see clearly through windows.

Source: MIT