An improved polynomials model using RBF network for fish-eye lens correction

With the rapid development of technology nowadays, large field of view is required in many occasions. Fish-eye lens boast of their extreme wide angle but the inherent severe distortion substantially limits their further development. In this paper, we proposed a polynomial approximating method based on fish-eye lens. Based on RBF Neural Network training, an imaging system model was established by obtaining the radial distortion parameters. Then it was simplified by Chebyshev Interpolation approach to reduce the amount of calculation in a large scale. The experiments show that an ideal rectification image can be acquired by RBF Network training. At the same time, the processing speed was accelerated by the multi-level method. The proposed method will be beneficial to the application of fish-eye lens in many fields. 1548-7741/Copyright © 2012 Binary Information Press.