Improvement 3-D Reconstruction Accuracy Considering Distortion in Stereovision Using a Set of Linear Spatial Filters

This paper presents a frequency-based stereovision approach to improve the accuracy of the three-dimensional (3D) reconstruction. In view of the fact that, stereo images are obtained from the different perspective position, so that each image can have the effect of projective distortion region (half occlusion, perspective foreshortening). These effects cause problems, since the projected distortion region cannot be estimated in terms of fixed-size block matching algorithm. An alternative approach is presented, which based on the output of linear spatial filters turned to a range of orientations and scales to make the correspondence analysis more reliable and robust. The responses of these filters at a given point constitute a vector called the filter response vector (FRY) is correlated instead of correlating area in the two images. In addition, an automatic procedure is used to evaluate and optimize the filters set by using the steering theorem and the singular value decomposition (SVD). The experimental result shows that remarkable improvement is obtained in the projective distortion region