In this paper, we propose a novel and efficient algorithm to reconstruct the 3D structure of a human face from one or a number of its 2D images with different poses. In our proposed algorithm, the rotation and translation process from a frontal-view face image to a non-frontal-view face image is at first formulated as a constrained independent component analysis (cICA) model. Then, the overcomplete ICA problem is converted into a normal ICA problem. The CANDIDE model is also employed to design a reference signal in our algorithm. Moreover, a model-integration method is proposed to improve the depth-estimation accuracy when multiple non-frontal-view face images are available. Experimental results on a real 3D face image database demonstrate the feasibility and efficiency of the proposed method.
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Conference||11th Pacific Rim Conference on Multimedia, PCM 2010|
|Period||21/09/10 → 24/09/10|
- Computer Science(all)
- Theoretical Computer Science