The Acceleration Algorithm Simulation for Atmospheric Turbulence Degraded Images Based on Kolmogorov- Arnold Network

Current atmospheric turbulence degradation image simulations struggle to meet the efficiency requirements of modern large-scale datasets. This paper applies Principal Component Analysis, Kolmogorov-Arnold Network and FiLM structures to simulation computations, proposing an accelerated algorithm for atmospheric turbulence degradation image simulation. Experimental results show that our algorithm achieves an average processing time of 3.08 seconds per image, which is significantly faster compared to the traditional phase screen segmentation algorithm (131.23 seconds per image) and the Zernike polynomials algorithm (18.14 seconds per image). The computational efficiency of the proposed algorithm is 42.61 times and 5.89 times that of the traditional methods, respectively. While accelerating the algorithm, only a 0.96% mean absolute percentage error is introduced.