This is a straightforward C++ port of the reference implementation of the Largest Triangle Three Buckets (LTTB) downsampling algorithm described in the paper "Downsampling time series for visual representation" by Sveinn Steinarsson. It is a single header, with a single class template that allows using different structures and data types.
Simply add the lttb.hpp header to your project. There are no binaries to install, and no dependencies outside the standard library.
Create a typedef to specify your time series datapoint type
#include "lttb.hpp"
struct ExamplePoint {
float x;
float y;
};
using PointLttb = LargestTriangleThreeBuckets<ExamplePoint, float, &ExamplePoint::x, &ExamplePoint::y>Then use the static method Downsample in the class. It can be used with iterators
std::vector<ExamplePoint> in = GetYourInputsFromSomewhere();
std::vector<ExamplePoint> out;
PointLttb::Downsample(in.begin(), in.size(), std::back_inserter(out), 50);...or pointers:
ExamplePoint in[500];
GetYourInputsFromSomewhere(in);
ExamplePoint out[50];
PointLttb::Downsample(in, 500, out, 50);