Capturing Cloth Geometry and Parameterization

This project presents a system to generate realistic cloth simulations by acquiring the geometry and a surface parameterization of real cloth from a captured video stream. Stereo processing with a technique for disparity map hole filling and smoothing is used to recover the geometry of the cloth. Image feature matching using SIFT features is combined with a surface interpolation method to determine a parameterization of the cloth surface. The resulting geometry from each frame is then re-textured using the parameterization to generate a virtual cloth simulation.

[1]  Volker Scholz,et al.  Cloth Motion from Optical Flow , 2004, VMV.

[2]  Igor Guskov,et al.  Efficient tracking of regular patterns on non-rigid geometry , 2002, Object recognition supported by user interaction for service robots.

[3]  Wolfgang Heidrich,et al.  Cloth Motion Capture , 2003, Comput. Graph. Forum.

[4]  Igor Guskov,et al.  Trackable surfaces , 2003, SCA '03.

[5]  Volker Scholz,et al.  Garment motion capture using color-coded patterns , 2005, SIGGRAPH '05.

[6]  Derek Bradley,et al.  Augmented reality on cloth with realistic illumination , 2007, Machine Vision and Applications.

[7]  D. Bradley,et al.  Augmenting Non-Rigid Objects with Realistic Lighting , 2004 .

[8]  David G. Lowe,et al.  Distinctive Image Features from Scale-Invariant Keypoints , 2004, International Journal of Computer Vision.