Develop point feature based 3D imaging system with application in finger imaging
Image feature based 3D model generation and in particular, structure from motion, the problem of estimating the 3D structure of a world scene making use of 2D images, is one of the most challenging and sought after research problems within computer vision. This can be mainly attributed not only to its academically challenging nature of the problem but also due to the abundance of applications in the field of medical, entertainment, architecture, military, commercial, biometrical and many other exciting avenues.
The goal of this thesis is to build an image baased 3D modeling system using point features and a single source with applications in biometrics. Two main problems that were successfully solved during building process are a) developed a new image stitching algorithm that is capable of handling rotation and scaling stitching concerns; b) developed an efficient 3D model generation method capable of handling thermal modality. Evaluation of the performance of the developed image stitching algorithms are also presented. Extensive computer simulations show that the proposed algorithm produces stitched images that give better results in terms of quality of the mosaicked image when compared to existing correlation (Pearson, Spearman, and Kendall correlation) mosaicking algorithms. The presented 3D model can be used for many graphical, robotic, medical, virtual reality and other applications.