This method attempts to correct the skew and perspective distortion that can occur in captured images. Typically, this distortion is common in images taken with a digital camera, though this method can be used with any image with similar distortion, regardless of the capture method.

The primary correction applied by this algorithm is to transform parallel text lines so they are horizontal, and to align column edges vertically. To do this, a certain number of characters, text lines and column edges are required. Column edges mean that the beginning and/or end of consecutive text lines have identical positions (i.e. the left edge of a column aligned left or both edges of a justified column). The minimal quantities are approximately 10 text lines and a single edge of two columns or both edges of one column. The accuracy of the algorithm increases when there are a good number of lines and column edges and when these are not confined to a small area of the image. If the image contains only figures or graphics, no transformation will occur.

Reference