Color Sequence Preserving Decolorization

Abstract

Many visualization techniques use images containing meaningful color sequences. If such images are converted to grayscale, the sequence is often distorted, compromising the information in the image. We preserve the significance of a color sequence during decolorization by mapping the colors from a source image to a grid in the CIELAB color space. We then identify the most significant hues, and thin the corresponding cells of the grid to approximate a curve in the color space, eliminating outliers using a weighted Laplacian eigenmap. This curve is then mapped to a monotonic sequence of gray levels. The saturation values of the resulting image are combined with the original intensity channels to restore details such as text. Our approach can also be used to recolor images containing color sequences, for instance for viewers with color‐deficient vision, or to interpolate between two images that use the same geometry and color sequence to present different data.

Examining Implicit Discretization in Spectral Schemes

“…As a spectral scheme with a linear lightness profile, the Kindlmann color map was chosen to facilitate comparisons between the other rainbow color maps and grayscale. We chose this color map over various other spiral color maps, such as the cubehelix [Gre11] or black body [Mor16] color maps, because it traverses a similar distribution of hue values to the traditional rainbow and jet color maps, it exhibits banding that cannot be explained by luminance variation [DPR∗18], and it has an established pattern of use by the visualization community [Mor16, STP17, YLL15, ZH16, DPR∗18]. As no accepted device‐independent definition of the Kindlmann color map currently exists, we reconstructed the color map directly from the original paper figure, modifying the lightness channel to ensure linearity in CIELCh.…”

Deep Colormap Extraction From Visualizations

“…In contrast, many existing colormaps order the colors based on their hues. Yoo et al [56] addressed the ordering problem by fitting a long curve in the CIELab color space to smoothly pass through the colors in a given visualization. Unfortunately, the method only works for continuous colormaps, not applicable to discrete ones.…”

“…Although the method can handle both continuous and discrete colormaps, it assumes the input visualization contains an explicit color legend. Yoo et al [56] found that only 29% of 611 web visualization images have a proper color legend. Image queries on Google search engine with keywords 'data visualization', and 'information visualization' also show similar results; see Supplementary Fig.…”

“…In summary, prior heuristic-based approaches suffer from tedious parameter settings, such as different numbers of discrete colors [8], and color order recovering [56]. To overcome the limitations, we develop a learning-based approach, where we prepare a dataset with careful considerations on the data distributions, chart types, and colormaps to supervise the network training.…”

“…The results benefit many usage scenarios, e.g., to optimize poorly-designed colormaps and to enable interactive overlays to improve readability. However, the method relies on detecting organized color legends in the visualizations, which may not be available, particularly for those found from the Internet [56]. Thus, some works suggested to directly extract colors from visualizations, e.g., by finding a long path of smoothlyvarying colors in Lab color space for recovering continuous colormaps [56], or by adapting k-means clustering to find discrete colormaps [8], [59].…”