Bernd Michaelis

3-d Surface Reconstruction in Spatial Frequency Domain Bernd Michaelis

The multiple view geometry in computer vision includes 3-D image and 3-D video processing techniques have received increasing interest due to the availability of high-end capturing, processing and rendering devices. Computer vision can define as a set of computational techniques aimed at estimating or making explicit the geometric and dynamic properties of the 3-D world from digital images. The 3-D world is projected onto 2-D acquisition surfaces. During this process, depth information can be lost. Furthermore, the classical 3-D measurement procedures based on photogrammetric causes systematic errors at strong curved surfaces or steps in surfaces (depth discontinuities problem). The challenges arise also from the foreshortening in one view of the stereo images, that occurs when the measured object have depth variation with the direction of the baseline. we demonstrate a theory of modeling the physical effects of perspective distortion in stereo vision systems. The proposed model is based on local spatial frequency representation and phase-difference-based technique as a widespread method for depth and optical flow estimation due to its superior performance.