Reconstruction of 3D model using three 2D images (looking at the same object in 3 different views)

I came across a requirement to reconstruct 3 dimensional image of an object using three cameras placed around it such that it (each camera's field of view together) covers the entire object under view.

• I have already worked a little on stereo vision (using two cameras). In this, we obtain a 3D point cloud. I don't think this will be necessary in my case since I need not calculate the exact x, y and z distances. I only have to get a 3D model. Is my approach right? I have read about Stereo Vision using multiple cameras. Should I use this?
• I came across this comment by Image Analyst: "images must actually undergo a filtered back projection reconstruction process to build the 3D image from the images taken at different angles." I did not understand this clearly. I do not have a stack of images, I have only 3 images of the same object but with different views. How should I proceed with this? What are the terms which are used in my case? Is it called 3D Reconstruction?
• I also saw about SfM (Structure from Motion). My object / scene is not in motion.
• I saw this code available. This too takes different slices of a brain and converts them into a 3D image. Is there anyway I can alter this code to suit my need or is it a completely different application?

As far as I have understood, with the three images (of different views of same object) I have to first match features (say SIFT) and then get the disparity (?). Or should I use Image stitching and deal with the stitched image (maybe create a 3D point cloud with this)?

Please tell me how to go about this application. What are the necessary topics I should read about?

My application is to only capture the object in 3D (using 3 cameras) and save this 3D image.

I have also come across Three-View Geometry (PART III of "Multiple View Geometry in Computer Vision by Hartley and Zisserman"). I don't think this is the same as what I want. In my case the scenes wouldn't overlap to a large extent. The field of view of each camera would overlap only a little (the cube's edges, as shown below), it won't be the same as in a stereo pair of cameras looking at a common scene. So stereo matching won't be feasible. Is my understanding correct? How should I deal with it?

[bhavisha parmar]

did you get the answer? did you finish that project? could you please help me for the same?