CSE 548 Computer Vision Final Project

This project is the final project of CSE 548 Computer Vision in Stony Brook University.

Author

• Junwei Jason Zhang
• Jian Jiang
• Zhenxiao Guo

Goal

Establish accurate one-to-one mapping between two faces using method combining 3D geometry information and 2D texture information.

Specifically, in the aspect of 2D image matching, we use non-rigid registration. In the paper of Efficient MRF Model for Non-rigid Image Matching, a non-rigid deformation method based on MRF is introduced and use TRW-S to minimize the MRF energy function. We follow this paper to finish the 2D image matching.

Algorithm

• Discrete Ricci Flow, which has been already finished with Prof. David Gu
• Markov Random Field
• ICM
• TRW-S to minimize MRF energy function