computer vision - How to identify changes in two images of same object -

i have 2 images know represent exact same object. in picture below, referred reference , match.

the image match can undergo following transformations compared reference:

1. the object may have changed appearance locally addition(e.g. dirt or lettering added side) or omission (side mirror has been taken out).

2. stretched or reduced in size horizontally (it not resized in vertical direction)

3. portions of reference image not present in match (shaded in red in reference image).

question: how can regions have "changed" in ways mentioned above identified ?

idea#1: dynamic time warping seems candidate once beginning , end of match image (numbered 1 , 3 in image) aligned corresponding columns in reference image, not sure how proceed.

idea#2: match sift features across images. tessellation produced feature point locations breaks image non-uniform tiles. use feature correspondences across images determine tiles match across images. use similarity measure figure out changes.

you might want consider iterative registration algorithm. want perform optimization find parameters of transform, in case horizontal scaling , horizontal translation. once optimize parameters have transformation between 2 images, transform 1 match other, , can use subtraction identify regions differences.

for registration take @ itk library. can gradient decent optimization using mutual information metric. has number of different transforms capture translation , scaling. code should run on sample images show.