Texture preservation and speckle reduction in optical coherence tomography using the shearlet-based total variation algorithm

Abstract

In optical coherence tomography (OCT), speckle residue, texture loss, and edge blur are the main factors that limit the image quality improvement during speckle reduction, especially for human retinal OCT images. This paper proposes a shearlet-based total variation de-speckling framework (STV method) to address this problem. This method includes a de-speckling regularisation term, an edge-preserving regularisation term, and a fidelity term. By assigning a suitable weight to each term, speckle reduction, texture protection, and edge preservation can be adjusted to an optimal match. Based on the split Bregman iteration, a fast numerical algorithm for this method is also proposed. We test the proposed method by applying it to 10 OCT images and comparing the results with those produced by the adaptive complex diffusion method and the Curvelet shrinkage method, both of which have been thoroughly confirmed as effective de-speckling methods. The performance of these methods is not only quantitatively evaluated in terms of distinguishability, contrast, smoothness, and edge sharpness with five quantifiable metrics, but also qualitatively analysed in terms of speckle reduction, texture protection, and edge preservation through four visual exhibition modes. Compared with the two representative methods, the merits of the proposed STV method in terms of speckle reduction and texture preservation have been confirmed by numerous experimental results.