Registration of biomedical images is a critical component of biomedical image analysis. The main contributions are as follows. In the paper, we propose a novel iterative closest point (ICP) method for 2D gel electrophoresis image alignment. The paper seeks to difine a novel similarity metric which is composed by Euclidean distance and information potential. The proposed metric combines intensity information of spots with geometric information of landmarks. The high accuracy and robustness of the algorithm indicate it is promising for gel image alignment. We firstly propose a registration method based on multi-objective optimization. The method optimizes multiple similarity metrics. It has been applied to solve the practical problem of large nonoverlapping field of view (FOV) between images pairs to be registered. Experimental results have shown that our method is very robust to initial transformation and noises. We propose a modified vector evaluated genetic algorithm and apply it to solve intra-modality image registration problem. Experimental results have shown the proposed method is more accurate than single-objective optimization.
修改评论