Novel view synthesis from two cartoon face drawings

We propose a framework for novel view synthesis for a cartoon face given two images of key views such as front and side views. Due to the structure inconsistency of cartoon faces in manually drawn images, matching region detection between different views is a challenging task. We propose a local complexity based segmentation that grabs facial components such as eyes, eyebrows and mouth that are not clearly separated by contour in cartoon faces. We automatically find the matching region in the two input key images and morph them to create intermediate views.

[1]  Frank Nielsen,et al.  Statistical region merging , 2004, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[2]  Frédo Durand,et al.  2.5D cartoon models , 2010, SIGGRAPH 2010.