Image and Video Mosaicing

Author : Jharna Majumdar

Abstract

The problem of Image and Video Mosaicing consists of stitching the selected frames of an image or video sequence by estimating the camera motion between the frames and thereby registering successive frames and creating the mosaic. The mosaicing techniques available for still images can directly be extended to video. An additional variable that is required for video is the time variable, which can be put to use effectively in different ways. Some of the common approaches which use this time variable effectively are Feature Tracking, Optical Flow and so on. Feature tracking approach tracks the features in successive frames thereby establishing a correspondence between the feature points, which in turn can be used to compute the motion between the frames. Optical Flow technique considers the brightness constancy equation while computing the motion between the frames.
In this paper, author introduces the basic concepts of mosaicing and describes image registration, a basic image processing technique, as the key challenging problem for the creation of mosaic. The author addresses feature based registration and optical flow based technique as two important methods for the creation of image and video mosaic.

Keywords—image registration, feature extraction, corner detectors, feature reduction, feature matching, coordinate transformation, optical flow, feature tracking