"Lasagne": a tool for interactive exploration of multivariate image data
- Peter Serocka (Universität Bielefeld, Bielefeld, Germany)
Abstract
As new technologies for gathing multivariate image data arise in many areas, the number of obtainable dimensions or layers per data set has increased dramatically. To quickly explore such data we visualize the similarity of all pixels in the data set with a given reference pixel. Two aspects are essential: first, providing a good similarity measure or metric which helps detecting relevant features that are not a-priori known. Second, computing this metric extremely fast for all pixels in the data set.
We present our tool Lasagne, which implements different metrics and delivers high interaction rates: The similarity image is generated in virtually the same instance as the user points to a pixel on the screen. This allows for visually scanning the data set in a contiguous way - a new experience in looking at multivariate images, revealing smallest details as well unexpected features.