Producing animated scenes from still images

An apparatus comprising:

• a processing unit;
• memory including an animation module, the animation module executable by the processing unit to:
• receive plural still images that are produced by an image capturing device at respective capture times;
• produce an animated scene by synthesizing the plural still images, wherein the synthesizing involves reordering the plural still images, wherein the producing the animated scene comprises:
• creating a graph based on an analysis of similarity among the plural still images;
• performing partial temporal order recovery to define a partial ordering among the plural still images by finding extremal paths in the graph, wherein an extremal path links a node under consideration to a node which is farthest from the node under consideration;
• extracting an output sequence from the plural still images using the partial ordering as a reference;
• and
• store the output sequence in the memory.

A computer-implemented method for producing an animated scene from plural still images, comprising:

• providing plural still images that are produced at respective capture times using a digital camera;
• decomposing a scene represented by the plural still images into a plurality of independent animated regions (IARs);
• creating a graph based on an analysis of similarity among the plural still images;
• performing partial temporal order recovery after creating the graph, wherein the partial temporal order recovery defines a partial ordering among the plural still images;
• extracting an output sequence from the plural still images, using the partial ordering as a reference, wherein:
• the creating, performing, and extracting are separately performed by a processing device with respect to the plurality of IARs,
• an IAR that has a level of motion above a defined threshold is further decomposed into plural interdependent semi-independent animated regions (SIARs), wherein the creating, performing, and extracting are independently performed with respect to at least a first SIAR, while the creating, performing, and extracting of other SIARs are performed based on the dynamics of a respective reference SIAR as a constraint;
• and
• storing the output sequence in a device memory.