Browse Prior Art Database

LUMINANCE PRE-DISTORTION FOR PERCEPTUAL COLORMAPS

IP.com Disclosure Number: IPCOM000238855D
Publication Date: 2014-Sep-22
Document File: 6 page(s) / 942K

Publishing Venue

The IP.com Prior Art Database

Abstract

Within the field of data visualization, there is a general problem of how to best choose colors to show numbers, and more particularly how to create color palettes that have a maximum number of hues and a near-linear brightness function. The point of data visualization is to convert data to pictures, so that people can see visual shapes and patterns that correspond to meaningful patterns in the data, such as clusters, outliers, or trends. However, typical schemes used to map data to pictures, particularly those schemes involving color, may introduce distortions and false edges, making it difficult to understand the form of the data. Also, for purposes of data visualization, color is typically considered an unordered visual element because the most salient aspect of color is hue (or frequency of reflected light), and ordering colors by frequency does not correspond to how color is perceived by humans. Effective use of color to show numbers is therefore challenging.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 31% of the total text.

Luminance Pre-Distortion For Perceptual Colormaps

Jonathan Helfman

Background

Within the field of data visualization, there is a general problem of how to best choose colors to show numbers, and more particularly how to create color palettes that have a maximum number of hues and a near-linear brightness function.  The point of data visualization is to convert data to pictures, so that people can see visual shapes and patterns that correspond to meaningful patterns in the data, such as clusters, outliers, or trends.  However, typical schemes used to map data to pictures, particularly those schemes involving color, may introduce distortions and false edges, making it difficult to understand the form of the data.  Also, for purposes of data visualization, color is typically considered an unordered visual element because the most salient aspect of color is hue (or frequency of reflected light), and ordering colors by frequency does not correspond to how color is perceived by humans.  Effective use of color to show numbers is therefore challenging.

Unlike hue, the brightness of colors has a natural ordering that corresponds to how humans perceive light.  Light may be easily perceived as ordered by brightness, either from light to dark or from dark to light, and the brightness variations may be easily interpreted by humans to understand visible forms and structures, even at low light levels where hues may not be visible.  Color brightness can be calculated by summing different ratios of the red, green, and blue components of a color.  When the brightness of each color in a palette (a ‘colormap’) is plotted in order and the resulting graph is a straight line, the colormap may be characterized as having a linear brightness function.  Colormaps with linear brightness functions, such as most grayscale colormaps, reveal form, structure, and patterns in data more effectively than colormaps with non-linear brightness functions which can introduce distortions.

However, most data visualization software packages default to using rainbow colormaps to show numbers, and therefore introduce distortions that obscure patterns in data.  Rainbow colormaps are rarely ordered by brightness, so they almost always obscure the form and shape of the data.  The three main sources of distortion are brightness distortion, color banding and false edges.  Also, in many cases colormaps for data visualization are calculated using colorspaces that have not been optimized for human perception, such as RGB (red, green blue) or HSV (hue, saturation, value).  These colorspaces correspond poorly to the human visual system.  In particular, equally spaced samples of these color spaces do not result in colors that look equally spaced.

In contrast, perceptual colorspaces such as CIELCH and CIELAB correspond much better to the human visual system.  Perceptual colorspaces are designed to model the major nonlinearities of how humans perceive color.  In particular, equally...