Images are taken by a DSLR camera, always set at a particular Focal length. The scene contains objects with varying distance to the cameras. Since the camera has undergone translation, homography no longer applies. Thus looked into Epipolar Geometry for pure translation. I know a very close estimate of the displacement of the camera (in between images, say Image 1 taken at x and Image 2 taken at x+d). Is there anyway to find the transformation that maps Img1 to 2 or vice versa to accomplish image registration (need to do image differencing after) ? Can the fundamental matrix and the displacement be used to transform an image to another ? I am using OpenCV for my implementation.
Without knowing a decomposition of the scene into objects and the depth of each object (distance to each object), it's not possible to achieve what you want. A very-close object will be affected much more by the translation than a very-distant object.