CS 766 Computer Vision, Spring 2016, Final Project

View Morphing - An Automated Implementation

Team Members : Ashish Shenoy & Rashi Jalan


What are we trying to do?

Given left and right image, create intermediate images that simulate the camera movement. We have also attempted at making the whole process automatic using existing feature selection methods.


Previous Work

Image Morphing (Beier and Neely, 1992)
1. Not shape preserving
2. There is distortion in the generated images
3. Unnatural.

 

View Interpolation (Chen and Williams, 1993)

1. Requires depth values
2. Creates arbitrary viewpoints

 


What is view morphing?

1. A shape preserving morphing algorithm.
2. Three step process:
    *Prewarp the left and right images to parallel views
    *Morph the prewarped images
    *Postwarp the interpolated morphed images.
3. Published in 1996 by Charles Dyer and Steven M Seitz.

 

 


Our Results

 

 


Manual view morphing with 30 feature lines



Manual view morphing with 50 feature lines



Automatic corresponding point selection using SIFT



Automatic corresponding point selection using SIFT



Automatic corresponding point selection using SIFT and automated Mesh Warping




For more details on the implementation and a better version of the result videos please refer to the prezi below





For the source code please contact Ashish Shenoy or Rashi Jalan and we will be more than happy to help!