A robust performance evaluation scheme for rectification algorithms in camera captured document images

Major issues faced by camera captured document analysis is to deal with the page curl and perspective distortions. Perspective distortion occurs when the world plane is not parallel to the imaging plane. This causes parallel lines to become non-parallel and the objects which are far away from the camera to seem smaller. Rectification is an unavoidable process on images with perspective distortion for precise representation. No standard measures for the performance evaluation of rectification algorithms exists with most of the evaluation done to concentrate in visual pleasing impressions. OCR based evaluation methods perform well for evaluation of rectification in text regions, but they fail to measure the performance of rectification algorithm on the non-text regions like mathematical equations, graphics, or tables. Feature based methods, on the other hand require manual assistance. An image based performance evaluation of rectification algorithms using SIFT(Scale Invariant Feature Transform) features was reported earlier. In this paper, a new method for performance evaluation using SURF(Speeded-Up Robust Features) features is proposed, which is more efficient and 12.07 times faster than SIFT based method according to the obtained results.