An efficient watermarking based integrity control system for medical images

摘要

Medical image applications like Picture Archiving and communication Systems (PACS) need transmission of images over insecure network. The images especially the Region of Interest (ROI) can be tampered while transmission due to the modifications such as JPEG, JPREG2K, scaling, rotation, filtering etc. As a result, this tampering will affect the disease diagnosis. Hence the verification of integrity of images and identification of the nature of modifications are important. In this paper we present a watermarking based integrity control system with an efficient classification of JPEG versus JPEG2K modifications. This system is proposed to detect the modification, identify whether it is global or local and classify globally modified images into different classes. It uses the cryptographic hash to verify integrity of the image. The image features namely the geometric moments, blockiness and the average of absolute difference between adjacent pixels within blocks are used for classification of modified images. The multi-classifier SVM that used for classification of modified images consist of binary SVM classifiers for each pair of modifications. The performance of binary SVM classifier for JPEG versus JPEG2K is improved by using these features. As a result the performance of multi-classifier is also improved.