V iQ oo
For the feature tracking tests, the initial film ofa
patient was converted to a digital image with the CID camera
and displayed on a monitor. The operator moved a cursor
(also displayed on the screen) through the image, choosing
and storing anatomic features of interest. Also, a
rectangular "patch" boundary was displayed on the monitor
(Figure 1).
Figure 1 - Feature Selection
The patch could be enlarged or reduced as necessary in order
to cover as much of the feature as the operator deemed
necessary to uniquely identify it. Each patch was stored as
a subset of the original image, along with the coordinates
of the patch boundaries and the coordinates of any
identified points of interest within the patch. These might
include the apex of a tooth root, the tip of an incisor, or
the center of the pituitary fossa (sella).
After the operator completed identifying points of interest
on the baseline film, any remaining films for the patient
were processed automatically. Each x-ray film was digitized
with the CID camera, and the resulting digital images were
Stored in the system and scanned by the software, tracking
the points already identified on the baseline image.
3. FEATURE BOUNDARY DETECTION AND TRANSFORMATION
Straightforward correlation methods match digital image sub-
pictures on the basis of pixel to pixel grey level
differences. However, if the features being matched differ
in shape, size, and orientation, the correlation method does
not work well. Several refinements to the basic correlation
- 165 -