[OPTICAL REVIEW Vol. 6, No. 4 (1999) 314-320]

Detection of Interval Changes from a Pair of Images: A Proposal of Pattern Histogram

Yucel UGURLU, Takashi OBI, Akira HASEGAWA, Masahiro YAMAGUCHI and Nagaaki OHYAMA

Imaging Science and Engineering Laboratory, Tokyo Institute of Technology, 4259, Nagatsuta, Midori-ku, Yokohama 226-8503 Japan

(Received January 7, 1999; Accepted May 7, 1999)

A new method to detect interval changes from a pair of images is presented, especially when the precise image registration is not available. We extend a well-known image histogram to multi-dimensions with some modifications and make use of it to elucidate differences between the images. The framework herein introduced is based on a comparison of the images in their multi-dimensional histogram space rather than in a spatial domain. Due to its flexibility, this approach is preferred as an alternative image comparison technique. The performance of the method is confirmed on simulated and real medical images and compared with the results from related conventional counterparts.

Key words : pattern histogram, k-means clustering, interval change, personal data, chest radiographs, pneumoconiosis.