An FFT-based algorithm for computation of Gabor transform with its application to cloud detection/classification

An efficient algorithm for the computation of 2D discrete Gabor transform is introduced. Under the assumption of non-overlapping windows, which is satisfied in many practical cases, the 2D Gabor coefficients can be calculated through an FFT-based scheme. Combining with the Kohonen self-organized map, this method is used in a real-world problem dealing with cloud detection/classification. Simulation results are also provided which show the promise of the proposed method.