Feature variation curve guided transfer function design for 3D medical image visualization

As an effective tool for revealing internal structures, direct volume rendering has been widely used in 3D medical image visualization. However, its quality strongly depends on the transfer function design in the rendering process of mapping data values to the optical properties. In this paper, a novel and intuitive method for transfer function design based on the feature variation curve is proposed. A set of 2D images are rendered for each adaptive sampling data value along uniformly distributed viewpoints in 3D space. Then the feature variation curve can be automatically obtained based on the analysis of extracted Fourier descriptors from the 2D images. By the guidance of the variation curve, the optimal transfer functions for revealing all the significant features inside 3D volume medical images can be easily generated and adjusted to fit the requirements of users. Additionally, GPU-based volume rendering makes our method can get a desirable speed. Compared with other methods for transfer function design, our method can obtain good volume rendering quality without requiring prior segmentation and tedious manual interactions. The experimental results demonstrate the effectiveness of our method.