A polynomial-division based correction model for camera calibration: a large comparative study

Abstract

These days, algorithms which enable correction of an image aberration such us barrel or pillow distortion can be found in a large number of scientific papers. Some sorts of algorithms have universal character, others can be used only with a certain type of optical systems. This paper compares several types of methods which were designed to remove an image distortion. The methods which are based on so called Polynomial Division Model (PDM) are only considered. On the grounds of extensive experimental results, a method which uses the polynomial division model was proposed. The method brought significantly better results in the correction process. It does not matter if the images are loaded with a pillow or a barrel distortion. The image correction calculation procedure is from a general point of view, very time demanding. The calculation with use of the PDM is fast, and does not need a large number of members of the correction polynomial. This rule is also valid for an image which is loaded with large distortion e.g., for Fish-Eye object-lenses - a newly proposed PDM preserves all good features of previous algorithms which are based on the PDM and extends them. All the results which are recorded in the experimental section show that the PDM is superior and more effective in comparison to other tested methods.