Appearance transfer techniques maintaining temporal coherence

A method implemented by a computing device, the method comprising:

• synthesizing a target region in frames of a target video by the computing device using frames in a video exemplar, the synthesizing including:
• identifying an emitter portion in a particular said frame of the target video being synthesized that is added to the target region from a previous said frame of the target video that occurs in a sequence of the target video before the particular said frame;
• and
• synthesizing color in the particular said frame from a corresponding said frame of the video exemplar as guided by a flow indicated in the warped previous said frame of the target video by reducing an effect of the identified emitted portion of the particular said frame.

A system comprising:

• an appearance transfer module implemented at least partially in hardware to synthesize frames in a target video using frames in a video exemplar, the appearance transfer module including:
• a warping module implemented at least partially in hardware to warp a previous said frame of the target video that occurs in a sequence of the target video before a particular said frame being synthesized;
• and
• a patch synthesis module implemented at least partially in hardware to synthesize texture in the particular said frame of the target video using a search and voting module to synthesize color in the particular said frame from a corresponding said frame of the video exemplar as guided by a flow indicated in the warped previous said frame of the target video.