Quality improvement Of 3-D voxel models based on histograms of model reprojection

This paper focuses on a method to improve the quality of the reconstruction of volumetric computer 3-D models calculated from multiple camera views. In particular, the idea of correcting some internal and external camera parameters, in order to obtain the estimated volumetric model with improved quality has been proposed. The authors proposed a solution which is built into volumetric reconstruction procedures. The proposed approach is based on an analysis of histograms of input images and their corresponding images of estimated model reprojection into the image planes. The proposed method utilizes cross-correlation to estimate corrections and characterizes itself as a non-optimization approach as opposed to the most common techniques. The effectiveness of the proposed approach has been tested on a widely used `Temple´ model from Middlebury dataset. The authors have added a known error values to the camera parameters to test the approach. The experimental results prove high effectiveness of proposed approach.