PCA and voxel similarity based 3D CT image registration accelerated by OpenCL
너칫투준 · 2012
Contents Abstract v ACKNOWLEDGEMENT vii List of Figures xi List of Tables xiv Chapter 1 1 INTRODUCTION 1 I. Background 1 II. Statement of problem 4 III. Related research works 9 III.1 Feature-based Medical Image Registration 9 III.2 Intensity-based Medical Image Registration 15 III.3 Accelerated Medical Image Registration 20 IV. Purpose and scope 21 Chapter 2 24 PROPOSED PCA AND VOXEL BASED REGISTRATION 24 I. Rigid Registration 24 II. Transformation Parameter Initialization 29 II.1. Binary volume extraction 30 II.2. Centroid calculation 30 II.3. Principal axes calculation 31 II.4. Parameter Extraction 33 III. Transformation parameter optimization 34 III.1 Similarity measure 35 III.2 Image transformation and Interpolation 36 III.3 Optimization 39 Chapter 3 41 PARALLELIZATION USING OPENCL 41 I. Parallelization using OpenCL 41 I.1 An overview of OpenCL 43 I.2 OpenCL implementation 47 Chapter 4 52 PERFORMANCE EVALUATIONS 52 I. Databases 52 II. Experimental results 57 II.1 Registration on phantom data sets 57 II.2 Registration on synthetic data set 60 II.3 Registration on patient data sets 62 II.4. OpenCL accelerated result 68 Chapter 5 73 CONCLUSIONS AND FUTURE WORKS 73 REFERENCES 75