Study and Analysis of Image Similarity Metricsand Image Filter for Vertebrae Pose Estimation

CHEN ZHENG(1751202)


X-ray image-assisted surgical navigation system has been broadly applied to spinal surgery. However, accurate path planning on the depth direction of X-ray image is difficult. This research studies 2D-3D image registration methods to obtain the optimum transformation matrix between X-ray and CT image frames. However the X-ray image is 2D and lack the depth makes the alignment very difficult. Three similarity measurement methods of 2D-3D image registration including Zero Normalized Cross-Correlation, Structural Similarity index and Mutual Information combined with intensity based optimization algorithm are applied to evaluate their performance in convergence, efficiency, and accuracy. Experimental results showed that the combination of Zero Normalized Cross-Correlation measurement method has the best performance of alignment between X-ray image and Digital Reconstructed Radiograph (DRR) images. Image filter is a common method for image pre-processing to reduce the noise and enhance the feature of image. The experiments proved image filter can influence and improve the accuracy of registration under the 0.5mm error target.