Noise in optical coherence tomographic (OCT) imaging degrades the quality of images, so improving image’s signal-to-noise ratio (SNR) is a focal problem in the development of OCT system. According to the distribution characteristics of signal and image noise collected by OCT system, a wavelet denoising method with a bivariate shrinkage function is improved, by increasing the radius of deadzone in a subband which includes more noise coefficient, to reduce noise. Some estimation parameters, such as contrast-to-noise ratio (CNR), SNR and edge preservation parameter, are used to estimate the improved effect of the image. The results show that it can reduce noise effectively and preserve edge information.