It seems a similar idea would be to adjust a linear model to a sequence of points and just maintain those that are outliers and other two that are end points on

It seems a similar idea would be to adjust a linear model to a sequence of points and just maintain those that are outliers and other two that are end points on the linear model.