Automatic threshold selection using the wavelet transform
CVGIP: Graphical Models and Image Processing
Artificial Vision for Mobile Robots: Stereo Vision and Multisensory Perception
Artificial Vision for Mobile Robots: Stereo Vision and Multisensory Perception
Multimodal Volume-Based Tumor Neurosurgery Planning in the Virtual Workbench
MICCAI '98 Proceedings of the First International Conference on Medical Image Computing and Computer-Assisted Intervention
Non-rigid Multimodal Image Registration Using Mutual Information
MICCAI '98 Proceedings of the First International Conference on Medical Image Computing and Computer-Assisted Intervention
Alignment by maximization of mutual information
ICCV '95 Proceedings of the Fifth International Conference on Computer Vision
Alignment by maximization of mutual information
Alignment by maximization of mutual information
Hi-index | 0.00 |
This paper describes a fast Mutual Information (MI) method for registering volumetric medical images. The new technique originates from the method designed by Viola [1] wherein registration is achieved by iteratively adjusting the relative position and orientation until the MI between two volumetric images is maximized. In this iterative process if n number of samples are used then there are O(n2) exponential calculations per iteration. The method proposed in this paper reduces the number of exponential computations by using an index table for estimating the Gaussian density functions (GDF). The index table is optimally pre-computed using automatic segmentation based on zero-crossing of wavelet transform. Thus a majority of exponential computations is reduced to index-intensity comparisons. The table lookup process is speeded up using a search mechanism based on probability priority. The proposed method has been successfully used to register both normal and pathological MRI and CT datasets. Experimental results show that this approach yields identical results in a fraction of time taken by the original method. The speedup increases with the number of samples used. For example, with 50 samples the speedup is 2.73 and for 100 samples it increases to 5.5.