Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for obtaining a first image of a document, the first image including first glare shown on a first portion of the document. Detecting, in the first image, the first glare shown on the first portion of the document. In response to detecting the first glare in the first image, obtaining a second image of the document, the second image including second glare shown on a second portion of the document. Generating, from the first image and the second image, a third image of the document that includes less glare than the first glare in the first image and less glare than the second glare in the second image.