A new algorithm developed by Adobe and Princeton University can automatically detect interferencefundamentals into photos and then edit them to delete them.
As researchers support in their work, they have been motivated to be able to click on an image editing tool such as Photoshop, iPhoto, or services like Google Photos or Instagram, to repaint bad pixels and bring them into their original form.
In addition, they wanted to find a way to automate the process so users do not need to open large desktop-based photo editing tools like Adobe Photoshop,
They did this by creating a new image analysis component called distractors, and then automatically using a computer algorithm called a distractor prediction model to automatically remove intentionally placed points. to distract from the images analyzed.
Circumcisers are generally the elements that stand out in a photograph, it is what drives the eye and focuses on it, that distracts you, so the algorithm that would recognize them was quite easy to create. It was difficult, however, to distinguish distractors from the general weaknesses of a photograph such as hyper-saturation or poor lighting.
To get the algorithm up to par for this project, the researchers who built it enlisted the help of Amazon's Mechanical Turk, which is an online service that allows people to randomly review or rate a business or an object.
Mechanical Turk users were asked to find splitters in a set of 1.073 photos, and then these data together with the data collected through an iPhone application called Fixel were edited by Adobe.
For this application, more than 5.000 photos were analyzed and recorded which parts of an image have been retouched.
Το τελικό αποτέλεσμα ήταν ένας αλγόριθμος που μπορεί να αφαιρέσει διαφόρων ειδών αποσπασμούς από τις εικόνες (πρόσωπα, cars, cut-off objects), but which also still needs a lot of work. But according to the researchers, the system "seems to promise something big".
Η full research work was conducted by Ohad Fried, a Princeton graduate student, Adam Finklelstein, Princeton professor, and Eli Shechtman and Dan Goldman from Adobe, and is available online with more examples. The document was also presented this summer at 2015 Computer Vision and Pattern Recognition Conference in Boston.
