3D Face Reconstruction From Single 2D Image Using Distinctive Features
H. M. Rehan Afzal,Suhuai Luo,M. Kamran Afzal,Gopal Chaudhary,Manju Khari,Sathish A. P. Kumar
DOI: https://doi.org/10.1109/access.2020.3028106
IF: 3.9
2020-01-01
IEEE Access
Abstract:3D face reconstruction is considered to be a useful computer vision tool, though it is difficult to build. This paper proposes a 3D face reconstruction method, which is easy to implement and computationally efficient. It takes a single 2D image as input, and gives 3D reconstructed images as output. Our method primarily consists of three main steps: feature extraction, depth calculation, and creation of a 3D image from the processed image using a Basel face model (BFM). First, the features of a single 2D image are extracted using a two-step process. Before distinctive-features extraction, a face must be detected to confirm whether one is present in the input image or not. For this purpose, facial features like eyes, nose, and mouth are extracted. Then, distinctive features are mined by using scale-invariant feature transform (SIFT), which will be used for 3D face reconstruction at a later stage. Second step comprises of depth calculation, to assign the image a third dimension. Multivariate Gaussian distribution helps to find the third dimension, which is further tuned using shading cues that are obtained by the shape from shading (SFS) technique. Thirdly, the data obtained from the above two steps will be used to create a 3D image using BFM. The proposed method does not rely on multiple images, lightening the computation burden. Experiments were carried out on different 2D images to validate the proposed method and compared its performance to those of the latest approaches. Experiment results demonstrate that the proposed method is time efficient and robust in nature, and it outperformed all of the tested methods in terms of detail recovery and accuracy.
computer science, information systems,telecommunications,engineering, electrical & electronic