Extension of the scatterplot approach to multiple images

The scatterplot is a tool for relating the content of an image to the content of another one. Its usefulness is limited to data sets composed of two (at most three) images. In this note, we show that the concept can be extended to series with a larger number of images, provided that the whole data set (N-dimensional space) is preliminary mapped into a R2 space. This can be done, for instance, by letting an “observer” compute its distance to any pixel. Scatterplots built from two such distance maps may help to visualize the content of the whole data set, to interpret it (number of homogeneous classes, …) and to proceed to the interactive or automatic segmentation (clustering) of the data.