Radon transform is a kind of technical method widely used in seismic data processing,but there are problems of low resolution and low computational efficiency in practical application.On the basis of phase shift hyperbola,the authors deduced time-invariant hyperbolic integral path,proposed sparse hyperbolic Radon transform based on Bayes principle,provided parameter selection principle of discrete transform,and introduced dynamic time windows to ensure the condition that reflection waves of different layers have the best convergence.Sparse hyperbolic Radon transform in the frequency domain has higher resolution and it can be realized in the frequency domain,so the computational efficiency can be drastically improved.Model data test and actual application result show that sparse hyperbolic Radon transform in the frequency domain can effectively remove regular interference,suppress random noise and improve the signal-to-noise ratio;in addition,it also has better fidelity and amplitude preservation.