Surface reconstruction and their similarities is a central problem in the field of vision based modeling. It arises in particular from the task of classifying and recognizing objects from their observed silhouette. Defining natural distances between image discontinuities creates a metric space of shapes, whose mathematical structure is inherently relevant to the classification task. One intriguing metric space is identified from using conformal mappings through Feature Extraction Algorithm of 2D stereo images. Surface reconstruction may be performed as unconstrained tracking and 3D data merging or as iterative structure from motion, or through constrained depth recovery using epipolar geometry. Immediate problems with these approaches are aperture problem, variation in stabilizing factor over time and object distortion at different viewpoints. The above said difficulties are avoided by introducing Feature extraction techniques. This approach results a denser depth map from the traits with variable window size to avoid distortion across the composite planar image. This constraint reduces the aperture problem during search. Wide-angle reconstruction of 3D scenes is conventionally achieved by extracting the features from stereo images, have been performed using two CCD Cameras under static and dynamic environment. The result shows the required optimal window requirement to get photorealistic surface reconstruction of physical environment scenes with minimum human intervention.

Key Words: Denser depth map, Feature Extraction, Surface reconstruction, Stereo Image, Optimal Window.