Suppose I have an ellipse with long axis $l_1$, short axis $s_1$ and say it has an orientation of $\frac{\pi}{2}$ (unnecessary but the graphic I remembered after I created the graphic attached that the usual convention is to have the long axis aligned to the x-axis).
Now suppose I apply an anisotropic scaling (scaling along two orthogonal axes); specifcally I scale by some factor $k_1$ in the direction of some vector $\vec{v}_1$ and by some other factor $k_2$ in the direction of a vector $\vec{v}_2$ that is orthogonal to $\vec{v}_1$. For non-trivial values such scaling maps an ellipse to an ellipse.
Given the directions of $\vec{v}_1$ and $\vec{v}_2$, the scaling factors $k_1$ and $k_2$ and the original ellipse parameters $l_1,s_1$ and $\theta_1$ (for completeness - its orientation relative to the x-axis / angle between long axis and x-axis), what are the new long and short axes - $l_1'$ and $s_1'$ respectively?
Initial attempt:
Assuming it to be centered at the origin, I applied the transformations to the ellipse parameterization and then sought out to transform that parameterization to some form $(l_1' \cos(t)\cos(\theta_1') - s_1' \sin(t)\cos(\theta_1'),l_1' \cos(t)\sin(\theta_1') + s_1' \sin(t)\sin(\theta_1'))$ where $\theta_1'$ is the angle $l_1'$ makes with the x-axis. This turns out to be nearly impossible or at least way above what I'm seeking, since there are so many variables involved.
I also thought of finding the max of the ellipse parameterization after applying the transformations, however also not exactly straight forward since it depends on the scaling factors and directions - but could be numerically achievable I suppose when one provides specific scaling factors and directions. However, I'm not seeking a numerical solution.
So is there a straightforward analytic way to find the long and short axis of the transformed ellipse?
Thanks in advance!
