Multi-channel Mutual Information Using Scale Space

We propose a new voxel similarity measure which utilises local image structure as well intensity information. Gaussian scale space derivatives are used to provide the structural information. Each derivative is assigned to an information channel of higher dimensional form of normalised mutual information. We illustrate the behaviour of the measure for a simulated signal and 2D medical brain images and demonstrate its potential for non-rigid, inter-subject registration of 3D thorax MR images as a proof of concept.

• Type: Conference
• Year: 2004
• Where: MICCAI
• Authors: Mark Holden, Lewis D. Griffin, Nadeem Saeed, Derek L. G. Hill
• Keywords: Gaussian Scale Space, Medical Brain Images, Medical Imaging, MICCAI 2004, Voxel Similarity Measure