We may be looking at an upcoming Photoshop feature

Sep 2, 2015 15:31 GMT  ·  By

A new algorithm developed by Adobe and Princeton University researchers can automatically detect distractions inside photos and edit them out.

As the researchers claim in their paper, their work was motivated by the one-click image enhancing solutions seen in tools like Photoshop, iPhoto, or services like Google Photos or Instagram, but instead of image filters and color corrections, they wanted to create a more complex one-click photo editing system.

Instead of users having to open large desktop-based photo editing toolkits like Adobe Photoshop, researchers looked to find a way to have the process automated.

This was done by creating a new photo analysis element called "distractors" and then have a computer algorithm named "distractor prediction model" remove distractors automatically from analyzed pictures.

Since distractors are generally elements that stand out in a photo, the computer model that looked for them was quite easy to set up. It was harder, though, to distinguish distractors from general photo flaws like over-saturation or bad lighting.

Humans were used to train the algorithm

To help train the algorithm for this task, the researchers first employed the help of Amazon's Mechanical Turk, which is a Web service that lets random people review or grade an operation or object.

Mechanical Turk users were asked to find distractions in a set of 1,073 photos, and this data was then compiled with the data collected through an iPhone app called Fixel, developed by Adobe.

This app analyzed over 5,000 photos and logged what parts of an image photographers have retouched while working.

The end result was an algorithm that can remove various types of distractions from images (faces, cars, cut-off objects), but which also still needs some work.

As the researchers put it, the system "shows great promise."

The full research paper, conducted by Ohad Fried, Princeton graduate student, Adam Finklelstein, Princeton professor, and Eli Shechtman and Dan Goldman from Adobe, is available online and contains more examples. The paper was also presented at the 2015 Computer Vision and Pattern Recognition Conference in Boston this summer.

The algorithm can detect and remove distractions from photos
The algorithm can detect and remove distractions from photos

Photo Gallery (2 Images)

New Adobe AI makes photobombing a lot harder
The algorithm can detect and remove distractions from photos
Open gallery