摘要
针对医学X线计算机断层(Computed Tomography,CT)图像,提出了一种基于尺度不变特征变换(Scale InvariantFeature Transform,SIFT)特征和近似最近邻算法的检索方法。首先通过SIFT算法得到图像的特征点和相应的特征向量,再采用近似最近邻算法进行SIFT特征向量的匹配搜索,得到数据库中与参考图像最相似的图像序列。实验结果表明,该法能检索到与目标图像细节相符的结果,大大提高了检索速度。与传统的基于纹理的检索方法相比,查准率和检索结果与目标图像的相似程度方面更佳,符合医学CT图像检索的要求。
A medical X-ray computed tomography（CT） image retrieval method is proposed by using Scale Invariant Feature Transform（SIFT） features and approximate nearest neighbor algorithm.First,the keypoints and corresponding feature vectors are extracted by SIFT algorithm.Then the approximate nearest neighbor algorithm is used to match the feature vectors.Finally the most similar images to the reference one are obtained.Experimental results show that the method can return the required images with good precision.Meanwhile,the approximate nearest neighbor algorithm increases the retrieval speed greatly.It has better performance than the conventional method based on textures,using the criteria of retrieval precision and the similarity between the retrieval results and the target image.It may meet the requirements of medical CT image retrieval.
出处
《生物医学工程学进展》
CAS
2011年第3期123-129,共7页
Progress in Biomedical Engineering
基金
复旦大学专用集成电路与系统国家重点实验室自主项目(09MS015)
关键词
图像检索
CT图像
SIFT特征
近似最近邻算法
image retrieval
CT images
SIFT features
approximate nearest neighbor algorithm
