Page 51 - 2025S

44 UEC Int’l Mini-Conference No.54

(a) Original im- (b) Vectorized (c) Vectorized
age with vector- background with- background using
ized background out face Data Hiding

Figure 5: Segmentation for image vectorization: (a) (a) Original image (b) Pre-processed image
original image with vectorized background, (b) vec-
torized background without face, and (c) vectorized Figure 6: Results of the preprocessing stage for em-
background using the technique ”Data Hiding in Bi- bedding: (a) original image and (b) pre-processed
nary Images with High Payload”. image with vectorized background and halftone wa-
termark.

posed of vertical white lines. This structured back-
ground, generated programmatically, replaces the and separability of the subject for further halftone
manual vectorization process and serves as a high- processing.
capacity region for embedding binary data using the This segmentation-based enhancement eliminates
extended ”Data Hiding in Binary Images with High the need for artificial vector outlines, which are of-
Payload” technique [7]. ten used to define regions of interest but can com-
This new strategy not only increases the embed- promise the natural appearance and diagnostic in-
ding bandwidth but also simplifies the workflow by tegrity of medical images. Instead, our approach re-
automating the generation of regions suitable for lies on a binary segmentation mask to isolate the tar-
high-density data hiding. An example of the pro- get region, which is then visually emphasized using
cessed image, including the segmented subject and a white-striped background pattern. This not only
the line-based background, is shown in Figure ??. ensures a clear and non-intrusive visual distinction
In order to optimize the embedding process and from the surrounding areas but also creates a struc-
avoid the visual artifacts introduced by artificial con- tured texture that is particularly well-suited for ro-
tours, we replace the previous vector-based outlining bust halftone-based data hiding. By integrating the
approach with a segmentation strategy that isolates enhancement directly within the segmented region,
the main object from the background. The process the method preserves the image’s semantic structure
begins with the conversion of the original RGB pa- while providing a reliable embedding area that sup-
tient photograph to grayscale. A high-threshold bi- ports watermark durability against compression and
narization then identifies background regions, which other common distortions.
are subsequently inverted and refined using morpho-
logical operations such as hole filling, area opening,
and closing with a disk-shaped structuring element. 3.2 Data Hider
This effectively isolates the patient’s silhouette.
Once the foreground mask is obtained, we gen- In our proposed method, clinical information is em-
erate an artificial background composed of vertical bedded into medical images using a reversible data
white lines spaced at regular intervals. These lines hiding technique that operates on uint16 DICOM
are inserted into the image outside the segmented images. The watermark—a binary halftone image
object by dilating the foreground mask and using combining a background of vertical lines with pa-
it to define a border area. The final image consists tient data—is encrypted using a logistic chaotic map
of the original subject preserved in its entirety and to enhance security and resilience against unautho-
a structured background that enhances the visibility rized recovery.