Color spanning Localized query

Let P be a set of n points and each of the points is colored with one of the k possible colors. We present efficient algorithms to pre-process P such that for a given query point q, we can quickly identify the smallest color spanning object of the desired type containing q. In this paper, we focus on (i) intervals, (ii) axis-parallel square, (iii) axis-parallel rectangle, (iv) equilateral triangle of fixed orientation and (v) circle, as our desired type of objects.

💡 Research Summary

The paper studies a family of “color‑spanning” geometric queries in a localized setting. Given a set P of n points in the plane, each point colored with one of k colors, the goal is to preprocess P so that for any query point q we can quickly report the smallest geometric object of a prescribed shape that (i) contains q and (ii) is color‑spanning, i.e., it contains at least one point of every color. Five shapes are considered: (i) intervals on a line, (ii) axis‑parallel squares, (iii) axis‑parallel rectangles, (iv) fixed‑orientation equilateral triangles, and (v) circles.

The authors first review prior work on computing the global minimum color‑spanning object for each shape. Those algorithms run in linear or near‑linear time but do not support queries for arbitrary q. The new contribution is a suite of data structures that, after a possibly expensive preprocessing phase, answer queries in polylogarithmic time.

For the one‑dimensional interval case (SCSI) the points are sorted, and for each point three lists are built: the smallest color‑spanning interval that starts at the point, ends at the point, or merely contains the point. With these lists a query reduces to a binary search for the two neighboring points around q, followed by a constant‑time comparison of the two candidate intervals (type‑I: q inside, type‑II: q on the boundary). The query time is O(log n) after O(n log n) preprocessing.

In two dimensions the approach is more involved. All minimal color‑spanning objects of the given shape are enumerated in advance. For squares (SCSS) the number of objects is Θ(nk); they are stored in a 4‑dimensional range tree where each node keeps the minimum side length among objects in its subtree. A query point q is processed in two ways:

• Type‑I: find the smallest pre‑computed square that already contains q. This is a 4‑D orthogonal range‑minimum query and costs O((log N·log log N)³), where N=Θ(nk).
• Type‑II: q lies on the boundary of the optimal square. The set of pre‑computed squares is split into nine sub‑categories according to whether they are “stabbed” by the vertical line through q, the horizontal line through q, or neither. Each sub‑category is handled by a dedicated auxiliary structure: three 3‑D range trees for stabbed squares (one for each side that may be extended) and six 2‑D structures that store L∞‑Voronoi diagrams of the relevant corners for non‑stabbed squares. These structures allow us to locate the square whose side can be extended to pass through q with minimal increase. The overall query time remains polylogarithmic.

Axis‑parallel rectangles (SCSR) are treated analogously. The number of minimal rectangles is Θ((n−k)²); they are indexed in the same 4‑D tree, and the same nine‑structure decomposition yields a query time of O((log N·log log N)³) with preprocessing O(N·(log N·log log N)²).

For fixed‑orientation equilateral triangles (SCST) the authors rely on known O(n log n) algorithms to list all minimal triangles, then index them in a 3‑D range tree. Queries are answered by the same type‑I/type‑II dichotomy, achieving logarithmic query time.

The circle case (SCSC) is more challenging because the exact minimum color‑spanning circle is hard to compute quickly. The paper therefore provides a (1+ε) approximation. The key idea is to restrict the circle’s center to lie on a set of pre‑selected lines. For a single line ℓ, the problem reduces to a one‑dimensional constrained version that can be solved optimally in O(n log n) time. By placing O(1/ε) such lines in a grid‑like fashion, any optimal circle can be approximated within factor (1+ε). The resulting data structure uses O(n) space, O(n log n) preprocessing, and answers queries in O(log n·log log n) time.

All five shape families share a common methodological framework:

1. Enumerate all minimal color‑spanning objects of the shape.
2. Store them in a high‑dimensional orthogonal range tree, augmenting each node with auxiliary information (minimum side length, maximum coordinate, L∞ Voronoi diagram of corners).
3. Separate candidate objects into those already containing the query point (type‑I) and those that must be expanded (type‑II).
4. For type‑II, further partition based on stabbing by the query’s vertical/horizontal lines, leading to a constant number of specialized sub‑structures that support fast “extension” queries.

The paper’s contributions are summarized in Table 1, showing preprocessing time, space, and query time for each shape. While preprocessing can be as high as O(n²·α(n)) for rectangles, the query times are uniformly polylogarithmic, making the approach suitable for applications where many queries are issued after a one‑time heavy preprocessing (e.g., geographic information systems, facility location with color constraints, or real‑time resource allocation).

In summary, the authors extend the classic color‑spanning problem to a localized query setting, devise efficient preprocessing schemes, and achieve logarithmic or near‑logarithmic query times for intervals, squares, rectangles, equilateral triangles, and approximate circles. This work bridges a gap between static optimization of color‑spanning objects and dynamic query answering, offering practical data structures for a variety of geometric optimization tasks.