Image Feature Based Method for Mail Retrieval

A mail retrieval method based on image feature is an innovative approach by which the postal department can realize the query of mail information. The paper chooses envelope images with handwritten address characters as retrieval object. Through a series of researches on a large number of image features, we select partition-block image feature, projection histogram feature and SIFT feature to construct feature vector. In order to accelerate processing speed, we optimize the retrieval procedure by matching feature component in multiple levels. The experimental results and comparative analysis have demonstrated that the precision of the proposed retrieval system reaches at least 99%, which proves the practicability and robustness of the method.