Analysis of Local Data Patterns by Local Adaptive Color Mapping

Abstract

Color, after position, is among the most effective visual variables to encode information. It is pre-attentively processed by the visual system, and if used appropriately, supports detection and correlation of patterns. Several global color mapping schemes (such as linear, non-linear and histogram-based) exist that support certain analysis tasks. However, static global schemes map data with a small local variation (within a data set of high variation) to small color differences. Often, these color differences are below the noticeable difference threshold of user perception or the display device. As a consequence, valuable information may be lost since data points or structures cannot be adequately perceived and correlations or other patterns of interest may be missed. Existing techniques to avoid this effect either require user interaction or are based on specific assumptions about the data. We introduce a novel automatic algorithm for local-adaptive color mapping that is applicable to dense data and is based on the idea to locally modify the color mapping to enhance the visibility of structures. This technique emphasizes patterns of interest within locally chosen color-ranges such that (1) the visibility of local differences is enhanced and (2) the introduced global distortion of the color mapping is kept small. This allows the perception of relevant patterns while approximately maintaining global comparability across the whole data set.