Visualization of high dimensional data through a single color image

Frédéric Blanchard, Michel Herbin et Philippe Vautrot

The visualization of multidimensional data is a crucial problem. The multidimensional images do not escape this difficulty. Most of the traditional techniques are failing in clearly displaying this kind of data. We propose a new visualization method by generating one color image to sum up the data set. The color allows humain eye to visualize the data set as a whole. The color image provides a global, synthetic and immediate visualization. The technique is based on a dimension reduction by Principal Components Analysis and a transformation of the first three principal components. Our technique is illustred by applications using different kinds of multidimensional data.