GENERALIZED PHOTOMETRIC STEREO FOR THE HYBRID REFLECTANCE MODEL USING AN IMAGE SEQUENCES

摘要

PURPOSE: A photo metric stereo method in a hybrid reflectance model is provided to estimate geometrical information such as surface albedo of an object, hybrid reflectance model parameters and surface normal vector from serial images to restore three-dimensional depth information of the object. CONSTITUTION: N images having different light source locations are received(S1). All of image pairs capable of being used for estimating image formation parameters are selected from the images, and the image formation parameters are repeatedly estimated from the image pairs(S2). Albedo and normal with respect to a diffused reflection component, which minimize diffused reflection error, are predicted from the selected image pairs(S3). A diffused reflection image is detected from the normal vector and diffused reflection albedo related with the diffused reflection component(S5). A total reflection image is detected from the diffused reflection image and entire image(S6). Image formation parameters related with total reflection are detected from the total reflection image(S7).