PROJECT TITLE :

Boosting Structure Consistency for Multispectral and Multimodal Image Registration

ABSTRACT:

In computer vision and computational photography, multispectral imaging is essential. It is vital to align spectral band pictures to avoid spectral information distortion owing to imaging device movement and alternation. The present multispectral data registration methods are robust but need a lot of computation. Sum of square differences (SSD) and sum of absolute differences (SAD) are both computationally efficient, however they perform badly on multispectral data. An SCB (structural similarity boosting) technique has been proposed to deal with this difficulty and to increase the structural similarity of multispectral pictures. For multispectral image registration, the usual measures can be used with SCB. It is possible to use the SCB transform to take advantage of the fact that local edge structures preserve relative saliency despite the nonlinear fluctuation between band pictures. Natural image statistical priors, such as the gradient-intensity correlation, are examined in order to develop a parametric SCB. The SCB transform outperforms current similarity enhancement techniques and outperforms state-of-the-art multispectral registration measures in experiments. SCB transform can be used to diverse multimodal data, such as flash/no-flash photos and Medical Imaging, because of the statistical prior.