George, If you don't want to save the full image data, then either poly data or unstructured data is the way to go. (Poly data will probably be a bit more efficient in ParaView, but either will work).
If you are looking for a format to store the data in, you might consider the VTK legacy data format. This format is documented in the VTK User's manual and here: http://www.paraview.org/Wiki/Image:VTK-File-Formats.pdf As far as the topology is concerned, you have a couple of options. One is just to define a bunch of "vertices" (0D cells). These will just appear as points on the screen. The advantage is that they are easy to specify: You just need coordinates and the data at that coordinate. The disadvantage is that they are just drawn as a series of points. You run the risk of holes appearing between the cells once the resolution of the data drops below the pixel level. (That might not be a concern for you.) The other option is to draw quadrilaterals either centered on each point or connecting 4 points. The quads will make sure that the space is filled but require you to specify more connectivity information. In either case, a Delaunay triangulation seems unnecessarily complicated. -Ken On 2/4/09 10:20 AM, "George Markomanolis" <[email protected]> wrote: Dear Ken, First of all thank you for your time. Unfortunately I don't want to save all the data to the files. For example I must save (and load, plot) 8GB of full image than 500mb If i "erase" useless points. I have no idea about the duration to plot 8GB and 500mb at paraview but I believe it's big (general the useless points are 85-95% of full image). I'll try of course to check it but we've developed a technique to "erase" useless data and I believe that it won't be accepted from the team except if it is fast with paraview. For example with gnuplot in order to have good quality I save data to eps format and after I convert it to jpg (plotting is done at the cluster and we download only the jpg files) but the convert is very low, for example 500mb may take 10-15 minutes to plot and save. I'll try to plot full image. For unstructured grid I don't know how to declare topology, I believe I need something like delaunay (triangulation) at matlab? Thanks, George Markomanolis On Wed, Feb 4, 2009 at 4:18 PM, Moreland, Kenneth <[email protected]> wrote: ParaView is capable of reading in a regular grid of data, mapping a color to it, and drawing it scaled on the screen. If your data is simply a raw array written to disk, then ParaView can read that in directly. Does that basically solve your problem? There are two ways to "erase" the useless points. The first way is to simply set the color map up so that these values are simply drawn in the background color. The second way is to use the "Threshold" filter to remove the uninteresting points. The Threshold filter will convert the data from a regular array to an unstructured grid. A warning about using Threshold though: the unstructured grid it converts data to is a less efficient representation so it is possible to actually require more data for the result if not a lot is removed. -Ken On 2/4/09 2:30 AM, "George Markomanolis" <[email protected] <http://[email protected]> > wrote: Dear all, I am newbie to Paraview and I want to ask you something. I am working with signals and I use gnuplot for plotting. Unfortunately it's slow for big signals. Our program is parallel so we can create files of many GBs. We do a tricky parallel plot, every cpu plots a part of the signal, otherwise we couldn't see the image from gnuplot (crash). So when I saw paraview I liked a lot but it isn't easy. I must explain what I want to plot: We use a techinque in order to cut the points that haven't energy. For example a specific signal with 12000 x 12000 mesh of points which is 8GB file it can go under of 1GB if we cut the useless points. So I give at the gnuplot only the points that I want to plot x,y,z and I use pm3d map because I want them in 2D not 3D for example see the image: http://www.markomanolis.com/files/plots/plot.jpg I would like to ask. Is this unstructured grid? I show an easy signal it could be with more random points. I have tried unstructured grid for 2 columns only and there is no surface between the columns. I used triangle strip and it was ok but I don't know if I could see the details like here (here I don't cut any useless point, in first image see wave details in the center): http://www.markomanolis.com/files/plots/62_0_0.jpg , http://www.markomanolis.com/files/plots/resFinal.jpg . Could I have these plots with paraview or it is good with more complicated plots? I must write a script/ program to convert gnuplot file to paraview but I am not sure about the topology, I must declare the topolgy for every point, right? I am confused because I have a lot of constrains for example if a point is alone then the topology is vertex if there is another point then line etc... Is there any way to plot this grid with something like image data. I want to give something like structured grid but NOT to give all the points (I don't need them). The last two images I sent you are with all the points for education propose. I want to plot something like first image but witho more complicated topology Thank you for your time, George Markomanolis **** Kenneth Moreland *** Sandia National Laboratories *********** *** *** *** email: [email protected] <http://[email protected]> ** *** ** phone: (505) 844-8919 *** web: http://www.cs.unm.edu/~kmorel <http://www.cs.unm.edu/%7Ekmorel> **** Kenneth Moreland *** Sandia National Laboratories *********** *** *** *** email: [email protected] ** *** ** phone: (505) 844-8919 *** web: http://www.cs.unm.edu/~kmorel
_______________________________________________ ParaView mailing list [email protected] http://www.paraview.org/mailman/listinfo/paraview
