针对医学图象数据库中大量存在的JPEG压缩图象，提出了一种新颖的特征提取算法。该算法提取的特征既包含空域局部信息又包含域频域局部信息，全面表达了图象的特征，使得基于内容查询的精度提高；压缩数据只需霍夫曼解码而不必恢复为原始图象即可提取特征，算法耗费时间大为减少；在JPEG压缩图象分辨率分析的基础上提出了由“粗”到“精”的查询。

In this paper, a new image feature retrieval method is proposed for large number of JPEG format medical images.The retrieved result, including both space domain and frequency domain information, expresses image feature perfectly and makes content based query more accuracy. The retrieving can be done just afterHuffman decoding but needn’t restore the compressed image data to its original state,thus the time consumed by the algorithml is greatly decreased. The method that the query can be done from“rough”to“fine”according to JPEGmulti-resolution analysis is also presented.