Two-dimensional bar charts

Abstract

Bar charts have been among the most commonly used types of visualizations throughout the years. They usually visualize a set of scalar data points by encoding the values into the length of the bars. The perception of differences of visual elements in length usually works well preattentively in human. Problems emerge when visual similarities present, due to small value differences or in the presence of big emtrema, causing the domain to stretch. Comparison of similar lengths has been proven to be more difficult when the distance between them increases. This thesis proposes a new approach to the problems. The concept is to turn the area limited by two side edges of a bar and the two auxiliary lines closest to its value into a grid. The cells of this grid are shaded basing on the ratio of value difference of the bar and the auxiliary lines. The cells can be divided into sub-grids to express finer values. The stacking of cells forms different shapes for different values, which human are good at detecting. The comparision task is thus performed better with this new technique. Further more, by counting and calculating the total value these cells represent, the value estimation process also achieves higher precision.