iat_dic is a wrapper of the dual inverse compositional algorithm [1] implemented in DIRT package by Adrien Bartoli (http://isit.u-clermont1.fr/~ab/Research/DIRT_v1p2.zip). DIC algorithm extends the inverse compositional scheme by Baker et al. (see the function iat_LucasKanadeIC) to jointly estimate a group of photometric transformations along with the warp. iat_dic returns only the geometric warp, while the photometric model accounts for any photometric distortions, thus making the alignment robust to intensity variations. The function relies on pixels around TEMPLATE edges to do the registration.

Input arguments:

IMAGE The image that must be warped in order to be similar to TEMPLATE
TEMPLATE The template (fixed) image
PAR A struct of parameters with fields:

PAR.iterations: the number of algorithm’s iteration (default:100). The algorithm may terminate earlier owing to a second criterion used by DIRT.

PAR.photoModel: the type of photometric transformation. Valid models: ‘GainAndBias’ (default for grayscale images), ‘SingleGainAndBias’ (default for color images), ‘MultipleGainsAndBiases’, ‘Affine’

PAR.transform: the type of geometric transformation. Valid strings: ‘euclidean’, ‘affine’, ‘homography’ (default)

PAR.initwarp: the initial transformation (default: identity transform)

PAR.maskBorder: the size of the TEMPLATE border that is masked (shrinkage of ROI) so that warped pixels of IMAGE are not mapped outside its borders (default value: 0)


Output arguments:

WARP Final estimation of geometric transform
MASK Binary image with TEMPLATE’s pixels used for the alignemnt.


[1] “A. Bartoli, Groupwise Geometric and Photometric Direct Image Registration, IEEE Trans. on PAMI, vol.30, no.12, 2008″


Back to Index