All "normal" images are projected in some way. So, I don't quite understand the words "normal" and "projected" in this context. However, there are some algorithms categorized under the keywords: "structure from motion". Perhaps they are what you are looking for. These slides give an overview and cite the important studies in this field: http://cs.nyu.edu/~fergus/teaching/vision/11_12_multiview.pdf
I join Sarper in being confused with what the word "normal" really means in this context. Additionally, my confusion is stretched even more when trying to infer volume from 2 images. However, with the help of properties of the devise that took the images and with some knowledge of physics, one can approximate the volume to certain extent (given the attenuation coefficients, be it x-ray, light, etc, have not changed).
Nevertheless, if you are talking about having an image and reconstructing it (minus the volume thingy) then you can look at this Radon-based example implemented in MATLAB:
For 3D reconstruction (from multiple projections), which is a step before calculating the volume, the filtered back project algorithm was and still widely used. One software which is freely available can be obtained from:
Normal I mean Ninety degree view to normal of the plane of the object (Object is rectangular box) and projected I mean any angle to normal of the plane of the object