3D Reconstruction of Human Head Based on Stereovision

The main purpose of the paper is to introduce an innovative way about 3D human head reconstruction and visualization based on stereovision. A new way of achieving geometrical reconstruction through the projecting two types of textures on head is advanced in this paper. The first type of texture is stripe, and the second has abundant texture. We adopt matching by using the second image by referencing to parallax distributing result obtained from the matching result of the first type of image. The actual texture reconstruction is one of a main content in 3D reconstruction. After creating TIN for the points on the head surface, the paper provides a new way about searching texture image and point´s coordinate on texture image by the outer spherical surface. Because of the influence of ray, optic and electric performance of camera, the grey or color in different image is not consistent, especially in the adjacent area. In order to keep the original information, we deal with the part of images in conjunct region and make the change of color meeting a certain function by referencing to the coordinate of outer spherical surface. Reconstruction of model´s head being used as the example, we capture the images in quite short time by tailor-made device and use the software developed by author to reconstruct the head triumphantly. A lot of experiments indicate the accuracy is better than 1 mm and the result of texture reconstruction is satisfying.