A Recursive Subdivision Technique for Sampling Multi-class Scatterplots

Sunspot Plots: Model‐based Structure Enhancement for Dense Scatter Plots

“…These approaches aim to reduce the density in highly occluded regions while preventing further decrease in density of sparse regions, which would result in information loss. Examples are best uniform sampling and non‐uniform sampling [BS06] which are both based on perceptual user studies, and multi‐class sampling [CCM∗14, CGZ∗19] that preserves point distributions. Ellis et al [EBD05] present a user‐controlled sampling lens enabling different sampling rates inside and outside the lens.…”

Pyramid-based Scatterplots Sampling for Progressive and Streaming Data Visualization

“…In contrast, continuous density fields bypass this issue and can effectively reveal patterns in highdensity regions. However, patterns in low-density regions (e.g., outliers) can become less visible [9,10]. For example, the region labeled by the purple dashed box is void in Fig.…”

“…By carefully choosing a subset of the data for display, well-designed sampling methods [5] can faithfully maintain the visual perception of relative data densities and outliers. Some recent methods [9,10] further attempt to preserve the relative class densities in multi-class scatterplots. All these methods, however, assume that the whole data is preloaded into memory, and thus they may not work well for progressive visualization, where the scatterplots are incrementally updated with new data coming.…”

“…All these methods, however, assume that the whole data is preloaded into memory, and thus they may not work well for progressive visualization, where the scatterplots are incrementally updated with new data coming. For example, the sampling results of successive frames generated by a KD-tree-based method [10] might not be temporally coherent; see the yellow lasso regions in the two frames shown in Fig. 1(d).…”

“…(c,d,e) The top and bottom rows show the results of different sampling methods in the 9th and 10th frames, respectively, where each result has around 1K points sampled from the original data chunk. Comparing (c) reservoir sampling [28], (d) KD-tree-based sampling [10], and (e) our progressive pyramid-based sampling, we can find our method more consistent in preserving high-density areas (see the LGA and JFK airports circled in green) and low-density areas (see the Expressway interstate 678 labeled by a red rectangle), while maintaining the density difference between different regions (see the Staten Island and the west bank of the Hudson River labeled by the purple and yellow dashed boxes).…”