基于特征融合和相关反馈的医学图像检索技术

目的 探索新型医学图像检索技术，以方便用户高效快速地查询到所需图像。方法 在分别研究基于颜色聚类、颜色纹理和形状等单一特征医学图像检索算法的基础上，提出了基于特征融合和相关反馈的医学图像检索算法。结果 通过对800幅标准临床内窥镜图像构成的数据库进行实验研究，分别给出了基于单一特征、特征融合以及相关反馈图像检索的查准率和查全率，并进行了比较分析。结论 利用特征融合和相关反馈逐步求精的方法，实现了更有效、更快捷的医学图像检索能力，

Retrieving Technique of Medical Images Based on Feature Fusion and Relevance Feedback

Objective To develop a new medical image retrieving technique in order to retrieve the desired images effectively and quickly. Method Bused on the study of single feature image retrieving techniques including color clustering, color texture and shape, a new algorithm with multi-feature fusion and relevance feedback was then proposed. Result By using 800 general clinical endoscopic images, the precisions and recalls under various conditions, such as single feature, multi-feature fusion and relevance feedback are finally given. Conclusion The proposed technique provides more accurate and flexible retrieving results due to its powerful information fusion and feasible man-machine interaction.