Video registration and image sequence stitching

A system for generating mosaic images from at least one set of sequential images of a scene, comprising using a computing device to perform steps for:

• receiving at least one set of sequential image frames of a scene;
• identifying the first and last image frames of each set of sequential image frames as a keyframe;
• extracting a set of feature points from each image;
• for each image frame, identifying feature point matches to an immediately preceding image;
• for each image frame, identifying feature point matches to an immediately preceding keyframe;
• for each image frame, estimating an amount of overlap with the immediately preceding keyframe and setting the frame as a new keyframe if the estimated amount of overlap is below a predetermined threshold;
• for each frame, identifying feature point matches to an immediately succeeding keyframe;
• for each keyframe, identifying feature point matches to every other keyframe;
• generating a match structure from all of the feature point matches;
• estimating optimal image registrations from the match structure;
• and
• constructing a static mosaic image from the estimated optimal image registrations.

A method for automatically generating a mosaic image from a sequential set of images of a scene, comprising:

• receiving set of images of a scene;
• evaluating each image as it is received to identify a set of feature points for each image;
• building a match structure of image correspondences as a function of the feature points identified for each image, said match structure including matches between image frames identified as keyframes, matches between groups of neighboring non-keyframe image frames, and matches between each non-keyframe image frame and the keyframe immediately preceding and succeeding each non-keyframe image frame;
• estimating optimal image registrations from the match structure;
• and
• constructing a static mosaic image from the set of images as a function of the estimated optimal image registrations.