A Directional Multiscale Approach for Speckle Reduction in Optical Coherence Tomography Images

This paper presents a new directional Bayesian despeckling technique for optical coherence tomography (OCT) images in the complex wavelet domain, which reduces speckle while preserving the detailed features and textural information. It has been shown that wavelet coefficients of natural images have significantly non-Gaussian statistics that are best described by families of heavy-tailed distributions. On the other hand, most of the edge information of layer boundaries in OCT images is located in the same direction. For these directional images, the use of heavy-tailed distributions does not seem to be appropriate for all wavelet decomposition subbands. So wavelet coefficients of the subbands which have almost the same orientation as the original image are modeled with heavy-tailed distributions such as the Cauchy, while the others are modeled with a simple Gaussian distribution. Within this framework, we design a maximum a posteriori estimator to remove speckle from noisy coefficients. Better results are obtained when we use the dual-tree complex wavelet transform which offers improved directional selectivity and near shift invariance property. Our results show that the proposed scheme outperforms some existing despeckling methods.