| Summary: | 碩士 === 國立東華大學 === 資訊工程學系 === 100 === A medical image is categorized into two regions: ROI (Region of Interest) and RONI (Region of Non-Interest). Generally, the ROI has the higher diagnosis value than RONI. Here, we deal with embedding the Electronic Patient Record (EPR) into the blood vessels images. To enhance the embedding capacity we embed data into both of ROI and RONI.
In this thesis, we introduce a new framework to enhance the embedding capacity of data hiding for medical image (DHMI). We apply the reversible data hiding (RDH) and non RDH (NRDH) on the region of interest (ROI) and the region of non-interest (RONI) in a medical image, respectively. We use the RDH (Histogram method) to embed the encrypted (AES128) EPR which includes patient’s private information into ROI. On the other hand, we use NRDH (LSB method) to embed the other metadata with lower security, e.g., the pure report of the image, into RONI for gaining larger embedding capacity.
Blood vessel image is tested, which is often used to detect blood vessel abnormalities such as aneurysms and atherosclerosis, and to evaluate blood flow. To designate the ROI in a blood vessel image, we cannot simply determine a shape to include a region as the ROI because the vascular system of human body is so complex. Unlike other organs have a specific shape, the sequence of blood flow capillaries are intertwined and have no fixed pattern. To cope with this problem, we propose a dynamic expansion to determine the ROI in a blood vessel image. Firstly, we find the edges of blood vessels using edge detector. Afterwards, we define the number of pixel expansions from the edge to broaden the blood vessels. Then, the original blood vessel and the expanded region are together zoned as ROI. The reason we use the expanded region as ROI is that the blood vessel abnormalities often occurs near the blood vessel. Finally, we can extract all of the data we embed in and the lossless image in ROI.
|
|---|
