A modality synthesis framework: Using patch based intensity histogram and Weber Local Descriptor features

Image modality synthesis has attracted much attention recently. In this paper, we propose a general framework for modality synthesis, which utilizes features such as the intensity histogram and the Weber Local Descriptor. A novel distance measure is proposed to search the most similar patch from the database. After identifying the corresponding patch in the database, modality synthesis becomes straightforward in that we just have to obtain the intensity from the corresponding patch of the targeting modality to fill in the corresponding location. Database reduction and a spatio population constraint are applied to confine the search space, thus reducing the computational time. We apply the proposed modality synthesis framework to a synthetic image study and an MRI synthesis application to demonstrate its effectiveness. A comparison with a state-of-the-art modality synthesis method shows that our method is superior in terms of synthesis quality.