A Novel 3D Image Processing Method Based on Spectral Layout

3D image processing is an important problem. Compared with 2D image processing, its difficulty is much higher due to the extra dimension. To fundamentally address this problem, we propose a novel method to embed the essential information in 3D image into 2D space via spectral layout. Specifically, we construct 3D adjacency graph to capture spatial structure of 3D voxel grid. Then we calculate the eigenvectors corresponding to the second and third smallest eigenvalues of its graph Laplacian and perform spectral layout to map voxels into pixels in 2D Cartesian coordinate plane. By using spectral method to capture the most critical structural information, the proposed method can achieve high quality 2D representations for 3D image, which enables to use 2D-based methods to process 3D images. The experimental results of visualization, clustering and classification demonstrate the effectiveness of our method.