Real–Time Correction of Cameras’ Geometric Distortions using GPGPU

Lens distortion correction is very important for machine vision applications. Estimation of internal camera parameters is necessary for precise 2D or 3D scene measurements. Nonlinear image warping operation needs a lot of computational effort and is critical for real–time systems. Modern GPGPU devices give the possibility processing such data but needs fast algorithm for calculations result. In this paper optimized coalescence algorithm for best GPGPU performance is proposed.