The Study of Fisheye Lens Distortion Model Estimation

Abstract

Fisheye lens have many characteristics, including short focal length and wide angle of vision. However, fisheye images also have great barrel distortion. That makes image recognition, measurement, and producing panoramic image in free software difficultly. This study focuses on the deformation of fisheye lens, using Canon EOS 5D camera and Sigma 15mm f / 2.8 fisheye len, taking the calibration pattern. A program was designed in order to establish more robust image process. The spline method and the kriging method (9 methods totally) are used to interpolate the deformation model of fisheye len, and bilinear interpolation is used to repair the image. The results show that the linear mode kriging method calibrates the deformation of the fisheye len better than other interpolation methods. The corrected image’s average offset value is 35.368 (pixels), and the image can be corrected 85.98% maximum error value.