Abstract In this paper we show how arbitrary surface reflectance properties can be incorporated into a shape-from-shading scheme, by using a Riemannian minimisation scheme to minimise the brightness error. We show that for face images an additional regularising constraint on the surface height function is all that is required to recover accurate face shape from single images, the only assumption being of a single light source of known direction. The method extends naturally to colour images, which add additional constraints to the problem. For our experimental evaluation we incorporate the Torrance and Sparrow surface reflectance model into our scheme and show how to solve for its parameters in conjunction with recovering a face shape estimate. We demonstrate that the method provides a realistic route to non-Lambertian shape-from-shading for both grayscale and colour face images.
A New Framework for Grayscale and Colour Non-lambertian Shape-from-Shading
Computer Vision – ACCV 2007 ; 11 ; 869-880
Lecture Notes in Computer Science ; 4844 , 11
2007-01-01
12 pages
Article/Chapter (Book)
Electronic Resource
English
Face Image , Surface Gradient , Photometric Stereo , View Synthesis , Radiance Function Computer Science , Image Processing and Computer Vision , Computer Imaging, Vision, Pattern Recognition and Graphics , Pattern Recognition , Artificial Intelligence (incl. Robotics) , Biometrics , Algorithm Analysis and Problem Complexity
Shape from Shading for Non-Lambertian Surfaces
British Library Conference Proceedings | 1994
|Recovering Shape by Shading and Stereo Under Lambertian Shading Model
British Library Online Contents | 2009
|Analysis and Approximation of Some Shape-from-Shading Models for Non-Lambertian Surfaces
British Library Online Contents | 2016
|Extraction of Shape Skeletons from Grayscale Images
British Library Online Contents | 1997
|