<p>In this paper, we present an approach for denosing and automatic dispersion compensation of optical coherence tomography (OCT) images based on deep learning (DL). Chromatic dispersion is a common problem that degrades the resolution in OCT images. We present an autoencoder based on transfer learning that depends on U-Net architecture for denoising and automated dispersion compensation of OCT images. The designed structure for dispersion compensation has an encoder-decoder pipeline. The autoencoder has been widely applied in dimensionality and noise reduction applications in image processing, and hence it is utilized in the proposed approach in this paper. The input for the designed structure consists of partially-compensated and noisy OCT B-scans. The output is composed of fully-compensated and denoised OCT B-scans with optimized retinal layers. To objectively evaluate the performance of the autoencoder and U-Net for OCT denoising and dispersion compensation, two metrics are adopted, namely peak signal-to-noise ratio (PSNR) and structural similarity index computed at multiple scales (MS-SSIM). The utilization of five input channels is adopted as the optimal configuration for autoencoder training, ensuring robust dispersion compensation and denoising of OCT images.</p>