Faster computation and more memory have opened up new avenues of research in image and video analysis. One method that strays particularly far from the classical bottom-up or geometric approach to computer vision is ``image similarity based image analysis''. This is a family of techniques which considers the image as a whole as the fundamental atom of analysis. A collection of images defines an image space, through a collection of techniques which map image similarities (however measured) between multiple images to a location of each image as a point in space. This gives new tools for image and video analysis.

Video Analysis: Wumap: A new video analysis method

Image Analysis: Isomap applied to image sets.