A new method to compute the distortion vector field from two images

In this paper, we present a new method to compute the distortion vector field from two images. The aim of this study is to characterize the distortion between to images. Several methods are based on optical flow techniques. However they are sensitive to image noise and grey-level variations between the two images. Our method is based on distance transforms. Images must be passed over twice before computing the distance transform and the distortion vector field simultaneously. Pixels are characterized by the displacement from the nearest feature element (surface for grey-level images). 3/spl times/3/spl times/3 chamfer masks are used for grey-level images. The relative weights of spatial versus grey-level displacements are considered. Experimental results are analyzed for both synthetic and natural images.