目的 针对现有角点检测算法的不足,提出结合Harris、Susan的混合角点检测算法,并应用于脑MR图像配准中.方法 首先通过Harris算子、Susan算子分别提取图像中Harris角点和Susan角点；然后对Harris角点和Susan角点取并集；通过引入两个加权因子ω1和ω2,分别对Harris角点响应值与Susan角点响应值进行加权计算,获得其角点强度,从而筛选出新的角点集合；通过归一化相关法和投票策略筛选出精确匹配的角点对；最后采用Powell算法进一步优化,获得图像最终配准参数值.结果 混合角点检测算法应用于脑MR图像配准能获得较高的配准精度和较好的稳定性.结论 相比于目前的角点检测算法,本文算法更适用于脑MR图像配准.%Objective To propose a hybrid corner detection algorithm by combining the Harris and Susan operators and applying it into brain MR image registration according to the brain MR image registration. Methods Firstly the Harris corner feature points and Susan corner feature points were extracted by using Harris and Susan operators. Secondly, the points and conducts weighted computation were merged based on two weight coefficients ω1 and ω2. After that, the feature points could be chosen further. Through normalization relevance method and voting mechanism, the final feature points were chosen further and matched between reference image and image needing registration. Finally, the Powell algorithm was used and the final transform coefficients were obtained. Results The experimental results show that this algorithm can be used for brain MR image registration and obtain higher registration precision and stability compared to the existing corner detection algorithms. Conclusion Compared to the existing corner detectors, this algorithm is more suitable for brain MR image registration.
展开▼
