Depth From Stereo Images

The aim of this project was to generate a 3-D visualization of a real world scene using a pair of Rectified Stereo Images. To demonstrate how the depth information contained in a pair of stereo images can be utilized to generate a 3-d view , satellite images of a hilly terrain taken from slightly displaced positions were used to determine the elevation or height variation in the terrain. The Symmetric Multi Window algorithm was used to perform stereo matching .Once the depth information was obtained from the matching procedure, this depth was used to obtain a perspective projection for the entire image.This perspective projection allows the user to specify his/her own view plane to view the image. A wireframe display of the projected image was implemented.By superimposing the color information from the image on the wireframe display a realistic visualization was achieved. The software was developed in Visual C++ 6.0 on Windows platform.

Presentation

Results

The images below are left and right views of a hilly terrain taken from a satellite
and the result of reconstructing the surface using stereo correlation is shown below it.

References:
1. Fusiello A Vito Roberto : Efficient Stereo with Multi Windowing
2. Foley J.D. A Van Dam et al: Computer Graphics Principles and Practice
3. Gonzalez C.R,R.E Woods : Digital Image Processing.