A Novel Image Registration Algorithm for Indoor and Built Environment Applications
Wei‐Yen Hsu · Computer-Aided Civil and Infrastructure Engineering · 2015
A novel registration algorithm is proposed for the applications of indoor and built environment in this study. A wavelet-based method is used to extract representative features, whereas an analytic differential approach is proposed to achieve fast registration of feature pairs. Finally, feature-based Levenberg-Marquardt method is applied to further refine registration results. This study presents the techniques used to process and merge indoor and built environment images to create and produce full scenes. The performance of the proposed method is evaluated by testing the images acquired by a hand-held digital still camera and in comparison with four registration techniques in terms of the squared sum of intensity differences and correlation coefficient. The results indicate that the proposed method is promising in registration accuracy, which are statistically significantly better than other four approaches. The use of such a system can assist the architects in facilitating the whole presentation of existing buildings and producing a complete overview and illustration of existing structures efficiently.