System and method for adaptive video fast forward using scene generative models

A computer-implemented process for automatically identifying similar image frames in one or more image sequences, comprising:

• acquire at least one image sequence, each image sequence having at least one image frame;
• select a query sample consisting of at least one image frame from one of the at least one image sequences;
• input a desired number of blobs to be modeled in generative model of the query sample;
• automatically learn a generative model from the query sample, wherein the generative model includes a set of model parameters that represent the desired number of blobs;
• and
• compare the frames in each image sequence to the generative model to determine a likelihood of each frame under the generative model.

A computer-readable medium having computer executable instructions for automatically determining a playback speed for an image sequence, comprising:

• selecting a query sample consisting at least one image frame;
• automatically learning a generative model from the query sample;
• providing at least one image sequence for playback;
• comparing the frames in the at least one image sequence provided for playback to the generative model to determine a likelihood of each frame under the generative model;
• and
• automatically varying a playback speed of the at least one image sequence in inverse proportion to the likelihood of each frame.