An image estimation method by header information in JPEG 2000 codestreams and its application to image identification

We propose an image estimation method by using header information in JPEG 2000 codestreams. Some information is directly extracted from header parts, and is used to estimate an original image without using any body data. From the estimated images, image signature values, which are utilized in MPEG 7 for image retrieval or image identification, are calculated as an image similarity index. Moreover, the relationship between estimated images and original ones are considered with the object of both image identification in the compressed domain and secure image management, by using the signature values. The experimental results show the effectiveness of the proposed method.