An approach is presented for multimedia similarity query using an on-line analysis of feedback sequence logs. The approach is based on user's feedback sequence accumulation and on-line collaborative filtering to predict the semantic correlation between the media objects in database and query sample. Edit distance is used to evaluate the similarity between current retrieval's feedback sequence and the prefixes of the records in the feedback logs. A prototype image retrieval system is implemented. Integrated with the retrieval method based on the generalized Euclidean distance, the performance of similarity query can be improved apparently. Experiments over 11 000 images demonstrate that this method outperforms the conventional ones.