Index: Full Text Index Contents: Conference Page 

Results

      We have selected some samples from the Visible Female dataset for our experiments to verify the quality of a multiresolution sampling procedure: fat-muscle, colon-transversal cut, colon inside cut, and colon-fat sub-images. The sampled images are shown in the following figures (Figure 1, Figure 2, Figure3). Considering that we did not apply pre-segmentation of visible human dataset before the sampling, the resulting images look very similar to the original images. Therefore, multiresolution sampling seems to be promising for further research. We have also presented our preliminary results about volume texture mapping of CT data set. First, we obtained some color textures from the Visible Female dataset. Then, we applied a simple classification without probability segmentation ([8]). In the last step, we applied the selected color texture and mapped it to corresponding tissue in the CT dataset. The results of the volume texture mapping are shown in Figure 4. As we see in this figure, the vessel is not corresponding to a real shape of the patient because the CT dataset does not provide enough contrast to distinguish vessels of fat or colon mucosa. The insufficient information of CT dataset is the main drawback for obtaining accurate images. Yet, we can explore Visible Human data to extract the physical laws that govern the shape formation of vessels. One of these laws is to minimize the size of the vessel tree and maximize the space filled by this tree to obtain a maximum distribution of oxygen. The problem is similar to finding the skeleton of an object in pattern recognition and computer vision. The difference between a vessel tree and skeleton graph is the relationship between parents and children.

Index: Full Text Index Contents: Conference Page