A spectral calibration technique, a data processing method and the importance of calibration and re-sampling methods for the spectral domain optical coherence tomography system were numerically studied, targeted to optical coherence tomography (OCT) signal processing implementation under graphics processing unit (GPU) architecture. Accurately, assigning the wavelength to each pixel of the detector is of paramount importance to obtain high quality images and increase signal to noise ratio (SNR). High quality imaging can be achieved by proper calibration methods, here performed by phase calibration and interpolation. SNR was assessed employing two approaches, single spectrum moving window averaging and consecutive spectra data averaging, to investigate the optimized method and factor for background noise reduction. It was demonstrated that the consecutive spectra averaging had better SNR performance.