We created a method to 3D-scan objects just using simple tools such as a regular webcam and projector. By using the projector to generate artificial features in the image, we bypass typical problems in stereo reconstruction involving insufficiently salient image features across views. We also demonstrate robustness of 3D stereo reconstruction from a pair of 2.5D images by reconstructing a human face from just two views spaced approximately 30 degrees apart. For such reconstruction with 2D images, the view angle spacing would have to be on the order of single digit degrees. Lastly we show the effectiveness of smoothing unstructured point clouds from a triangulation by generating very realistic looking faces from undersampled point clouds.

Built With

Share this project:

Updates