Micro image matching with grouped features

In this paper, we propose a micro image matching technique with grouped features to facilitate navigation under microscope. The features in micro images are detected and grouped as L-shapes first, then these features are matched across multiple images, after that, we estimate and optimize the transformation between every two consecutive images to represent the structure of motion. Finally, all the images in the sequence are fused according to the optimized transformation to build a Euclidean map of micro object under the microscope. This map is further fused into the perspective view image to assist the operator in finding the interested region more efficiently and accurately from the microscope