Rapid Multimodal Image Registration Based on the Local Edge Histogram

Due to significant differences in imaging mechanisms between multimodal images, registration methods have difficulty in achieving the ideal effect in terms of time consumption and matching precision. Therefore, this paper puts forward a rapid and robust method for multimodal image registration by exploiting local edge information. The method is based on the framework of SURF and can simultaneously achieve real time and accuracy. Due to the unpredictability of multimodal images’ textures, the local edge descriptor is built based on the edge histogram of neighborhood around keypoints. Moreover, in order to increase the robustness of the whole algorithm and maintain the SURF’s fast characteristic, saliency assessment of keypoints and the concept of self-similar factor are presented and introduced. Experimental results show that the proposed method achieves higher precision and consumes less time than other multimodality registration methods. In addition, the robustness and stability of the method are also demonstrated in the presence of image blurring, rotation, noise, and luminance variations.

Mattes’ Mutual Information Metric for Multimodality Registration of DESS and T2 Mapped Knee Articular MR Sequences

Registration of multiple MR sequences remains a challenging problem. The Insight Toolkit (ITK) implements the Mattes’ mutual information metric for multimodality registration. Here, example source code, data, and executable files to implement the Mattes’ mutual information metric in ITK are provided. Multiple MR sequences of the knee are used as example images. This serves as a companion manuscript for a permanent archive of the source code, executable file and example data and results.