one to look at the 3D dataset as a whole. Disadvantages are the 
difficult interpretation of the cloudy interiors and long time, 
compared to surface rendering, needed to perform volume 
rendering., (www.cc.gatech.edu, 2001). In our software, colour 
and opacity classification can be done interactively by viewing 
the changes of the appearance of the volume. Furthermore, 
gradient opacity function can be used for classification too. 
Volume rendering can be implemented by various methods. We 
used ray-casting methods in this study. 
2.2.1. Ray Casting 
For every pixel in the output image, a ray is sent into the data 
volume. According to the predefined sampling interval, along 
the ray the colour and opacity values are obtained by interpola- 
tion. The interpolated colours and opacities are merged with 
each other and with background by compositing in back to front 
order to yield the colour of the pixel. These compositing calcu- 
lations are simple linear transformations, (Schroeder, et.al, 
1998). This technique is called composite ray casting. 
During ray casting, to find the colour and opacity of a pixel, 
instead of compositing, maximum intensity values or average 
intensity values can be used as colour of pixels. This means 
that, when a ray travels along the data set, if the final pixel 
colour is assigned to the maximum density value along the ray, 
this is called maximum intensity projection (MIP). If the final 
colour of the pixel is computed by averaging the densities of 
pixels along the ray, this is called average intensity projection 
(AIP). Rendering with MIP and AIP, on final image some 
intuitive interpretations are needed. Because, the location of the 
maximum or the average values aren’t known. In this case it is 
not possible which objects are behind another. 
3. SEGMENTATION OF CT AND MR IMAGES 
Segmentation is the process of classifying pixels in an image or 
volume. It is one of the most difficult task in the visualization 
processes. For reconstruction of medical 3D surface and vol- 
ume, interest tissue boundaries should be distinguished from 
others on all the image slices. After the boundaries have been 
found, the pixels, which constitute the tissue, can be assigned to 
a constant grey level value. This constant value represents only 
this tissue. 
Label values can be used as isocontour value for surface render- 
ing. For volumetric rendering, in spite of the surface properties 
of the tissues, their inner properties are also important. Because 
of this reason, we should find the opacity values for the indi- 
vidual voxels. And near this, wee need different colours to 
separate the volume the elements which belong to different 
tissues. For these purposes, segmentation results are used too. 
In this study, we have used 3 different segmentation ap- 
proaches. These are; interactive histogram thresholding, contour 
segmentation and manual segmentation. 
3.1. Interactive Histogram Thresholding 
The simplest way of image segmentation is thresholding. By 
this technique, according to image histogram, possible threshold 
values are found. The pixels, that have values above or below 
this threshold is assigned to constant values. Thus a binary- 
segmented image is obtained. One can choose more than one 
thresholds. In this case, between values of these thresholds are 
replaced with constant label values. In our software, in spite of 
histogram analysis, we have also presented an interactive 
thresholding option. By this option, when one changes the 
threshold by using a track bar, its effect is seen on the screen 
synchronously. By the time user decide that the optimal seg- 
mentation has obtained, he/she can change the threshold. After 
thresholding, on the images, there would be many holes and 
also many small areas. To delete, unwanted small areas, we 
make a connectivity analysis, (Gonzalez, 1987, Teuber, 1993). 
By this analysis, the areas that are smaller than the area thresh- 
old are deleted. After connectivity analysis, still there might be 
some unwanted pixels on the image. We have written functions 
to delete these areas manually. After thresholded segmented 
regions have been obtained by using morphological operators 
such as erode/dilade, we fill or delete the remaining holes. The 
final segmentation is recorded as a file. 
3.2. Contour Segmentation 
By this method, possible boundary value of a tissue is selected 
with histogram analysis. This value is assumed to be the con- 
tour value and the interested image is contoured by tracking this 
value. After contouring, small areas can be detected automati- 
cally by connectivity analysis or manually with hand. After 
refinement of contours, we assign labels to pixels which are 
bounded by the counter lines. If user doesn't like the contouring 
result, he/she can ignore it and make a new segmentation easily. 
3.3. Manual Segmentation 
With the automatic segmentation procedures, it is inevitable to 
make some incorrect label assignment. So in the literature, 
manual segmentation is said to be still the best method. For 
precise medical applications, manual segmentation will give the 
best results. In this case, user draws the boundaries of the inter- 
ested region by using mouse pointer. User can make editing 
during manual segmentation. However, manual segmentation is 
too time consuming. It can take hours or sometimes days to 
segment complex MR images by manual segmentation. 
4. EXTERNAL FACE SURFACE CONSTRUCTION 
with DIGITAL PHOTOGRAMMETRY 
In this study, we have written a photogrammetric software 
module for external face reconstruction. We have tested the 
module for small objects and have good results. But because of 
speed limitations of our computer we couldn't test it for human 
head. We took photographs of the patient head with a mask on 
the head with the control points from multistations. We have 
calibrated and oriented pictures with the software by using 
bundle block adjustment with 10 additional calibration parame- 
ters. After orientation of the pictures, we begin the automatic 
matching procedure to measure the face surface points. For this 
purpose, we have written adaptive least squares matching 
(ALSM) with epipolar constraint. Firstly, the interest areas are 
pointed with a window on all images. Then our program, subdi- 
vides this window in to small windows on the master image 
during matching procedure automatically. For epipolar con- 
straint, at the start of the procedure user enters minimum and 
maximum Z values for the interest area. These values can be 
obtained approximately from priori information. Then the pro- 
gram computes minimum and maximum Z values for each sub 
window when processing them. And by using these Z values 
and exterior orientation parameters of the images, it finds the 
epipolar lines on the search images. Then along these lines, a 
cross-correlation matching is performed and the pixel where the 
—252- 
  
  
A PA no m