national museum of roman art cad download

It shows, for every mesh, the mean distance between the nodes defined in M ′ by the filling algorithm and their related nodes in M.

Table 1 summarizes the numerical results obtained. Figure 7 illustrates a picture of the meshes with holes we have been working with. To prove this point, the holes have also been filled using a commercial software and the results have been included in the comparison. The experimentation tries to elucidate not only if it is possible to fill holes on a 3D surface by means of image restoration algorithms, but also the usefulness of this methodology. The meshes have been damaged with a number of holes, from 2 up to 6, all with significant sizes and different kinds of surfaces (smooth surfaces, surfaces with sharp edges, flat surfaces, etc.), that have been restored by using the algorithms and. The method has been tested with a database of partial views from 20 freeform objects sensed with a Minolta VI- 300 scanner. Obviously, this distance is 0 except for the nodes that fit in the filled holes. The comparison is based in the measurement of the Euclidean distance between the nodes belonging to M ′ and their related nodes of M. This verification is made by comparing the data of the original mesh, M, with the data of the repaired one, M ′.

Once the image has been restored and the repaired holes have undergone the inverse transformation 2D to 3D to be merged with the defective mesh and provide the repaired M ′, it is time to verify the results. In all of them, the maxima are easily identify. In figure 6 the graphics for three different meshes are shown. This value is α = 0.87, and it is the number that has been used to generate the image that will be repaired with the image restoration algorithms. The value of α that causes a maximum of O I h has been obtained after a statistical analysis of the results provided by all the meshes we have worked with. To make this calculation independent of the mesh characteristics, q I h has been defined as proportional to the mesh edge average length, d : I h m (5) where α is the proportional factor and represents the variable related to O I h. To elucidate the best value for the size of the pixel (that is equivalent to the value of the image resolution), O I h has been calculated with several values of q I h for every mesh used in the experimentation. The value of the resolution that must be chosen is what supplies a maximum for O I h.

From this point, O I h is decreased again, since the number of pixels in the image is increased while the number of optimally occupied pixels is not. As resolution is increased, the number of optimally occupied pixels is, and so does O I h up to a value for which every pixel is optimally taken up. way, I h must be a small value for small image resolutions.

National museum of roman art cad download