I wish to create a Projection of Image. As in I have an outline of foot. Which is a binary image. I wish to create a 3D projection of it.
I have side view and foot outline. If I project the two and try to get the intersecting area I would get a rough foot model.
Can anyone help me with some codes or references.
Any Help would be highly appreciated.
I am not sure if I am getting your question right; would you like to transform a 3D model to a 2D model? something like a mapping?
And if this is the case, then perhaps you are looking for a function at "viewmtx" at http://www.mathworks.com/help/matlab/ref/viewmtx.html;jsessionid=80c0f8a85f221af448eea40bd7da (I actually got this form, hopefully, a similar question at http://www.mathworks.com/matlabcentral/answers/9523-2d-projection-in-a-3d-plot)
....In the course of trying to understand the question, i have attached some figures to see if they reflect what you are asking; please check them to see if the figures reflect the question.
x = imread('foo.bmp');% its 2D image
> xx = double(x);
> y = imagesc(xx)% gives 2D image
> z = mesh(xx)% gives the pixel values as 3rd dimension.
> Enjoy it
image3D = cat(3, slice1, slice2, slice3, slice4, slice5);
You could put it into a loop
for slice = 1 : totalNumberOfSlices
thisSlice = GetSlice(); % Whatever you have to do to get one 2D image.
if slice == 1
image3D = thisSlice;
else
image3D = cat(3, image3D, thisSlice);
end
end
Sam Its not slices da. The thing is I have just 2 views Side view and top view. First of all I need to create 1 3 Dimensional box grater than the size of both the views. Next what i need to do is like back projection. Where in say whatever is the image is top view is projected throughout the length of the box. And similarly for the side view. The intersecting area will give a rough shape of the object. Did you get the problem now? I have attached an image which is self explanatory. Just check it.
Got it...i thought it as two slices sorry..if i get i will let u know.
I think it's hard to construct the 3d model from only these 2 views you've indicated. A more common way is to use rectified stereo images to compute the depth information which can be utilized to reconstruct a 3d representation.
But there are also many other algorithms which do not expect those Stereo images (e.g Structure from motion).
Basically it boiles down to finding one and the same keypoint of the object on several captures. If you can't even see one and the same point on both recordings its hard to predict the object shape. That is why in photogrammetry, if you want to create a 3D image from 2d recordings you would use many different captures, from various angles (like 15 and more and not only 2).
There are many ressources and also some implementations available to perform this task. However if you are really limited to a top and a front view and dont even have stereo images of those I would believe this to be nearly impossible (but i may be wrong ;) ).
You can find a lot more information with the keywords: Stereovision, 3D reconstruction, photogrammetry, 2D to 3D conversion.
Regarding Stereo-Vision, there is an implementaion available in openCV.
See also the attached wiki article.
http://en.wikipedia.org/wiki/3D_reconstruction_from_multiple_images
@Parth From your description, one technique you might want to check is the "voxel coloring" (S. Seitz) it is of course a stereo vision techniques which usually requires as Gordon said more than 2 views of your object/scene if you hope to achieve some nice and dense reconstruction.
- Badges
- Science topic