Reconstruction algorithm for optical coherence tomography without requiring k-domain resampling

Axial resolution and speed are critical system parameters in optical coherence tomography (OCT). In conventional image reconstruction, the fast Fourier transform (FFT) and k-domain resampling are required to improve the axial resolution, which would get worse as the depth increases. In this study, a linear relationship model between interference spectral frequency and depth was established, and support vector regression (SVR) was employed for spectral fitting. The sparrow search algorithm (SSA) was used to optimize hyperparameters, while partial least squares (PLS) further refined the predictions to achieve more accurate reconstruction. To demonstrate the effectiveness of our method, the axial resolution and SNR were compared with FFT, spectral-estimation optical coherence tomography (SE-OCT), autoregressive (AR), and iterative adaptive approach (IAA) techniques. Experimental results demonstrate that, compared with the conventional FFT method, the proposed approach improves axial resolution by a factor of two and resolves structures visible only under full-bandwidth conditions. In addition, the method provides consistent contrast and speckle statistical characteristics. Through a hierarchical search strategy, the computational burden of spectral fitting is substantially reduced. Benefiting from the elimination of spectral resampling and requiring only a single spectral fitting step, the B-scan computation time can be shortened to the order of tens of milliseconds. These results indicate that the proposed method has strong potential for fast, high-resolution OCT imaging, offering a feasible pathway toward real-time biomedical imaging and large-scale deployment of super-resolution OCT.