Photorealistic Object Reconstruction Using Perturbation Functions and Features of Passive Stereo Projection

摘要

An effective approach to the calculation of volumetric models based on images of silhouettes in the object space is described. A method of forming the contours of a real object from a sequence of images is proposed. The construction of a bounding pyramid using the projection of the center and silhouette is shown. The intersection of all bounding pyramids determines the geometric representation of the object. Methods of obtaining information about the depth of the scene are proposed. A method of binocular stereo reconstruction of three-dimensional objects based on perturbation functions is proposed. In the method, interference is linear and requires linear compensation methods that are performed using the GPU by means of standard OpenGL. Defines the portion of the frame buffer (channel) in the GPU in which the image is output. Passive stereo uses two channels that process images intended for the left and right eye. Both channels handle a single main scene graph representing a user-defined model and attached to a root node (scene-root). Channels differ in the settings of the projection transformation (view-transform), which is determined by known parameters, such as the size and distance to the screen, the position of the user’s point of view, stereo base and others, which are specified in the calibration file. The paper shows the advantages of the proposed method over the known approaches of reconstruction of three-dimensional models.