1: \begin{abstract}

2: Given a compact semi-algebraic set $\S \subset \R^n$ and a polynomial map $\f : \R^n \to \R^m$, we consider the problem of approximating the image set $\F=\f(\S) \subset \R^m$. This includes in particular the projection of $\S$ on $\R^m$ for $n \geq m$.

3: Assuming that $\F \subset \B$, with $\B \subset \R^m$ being a ``simple'' set (e.g. a box or a ball), we provide two methods to compute certified outer approximations of $\F$. 

4: Method 1 exploits the fact that  $\F$ can be defined with an existential quantifier, while 

5: Method 2 computes approximations of the support of image measures. 

6: The two methods output a sequence of superlevel sets defined with a single polynomial that yield explicit outer approximations of $\F$. Finding the coefficients of this polynomial boils down to computing an optimal solution of a convex semidefinite program. We provide guarantees of strong convergence to $\F$ in $L_1$ norm on $\B$, when the degree of the polynomial approximation tends to infinity. Several examples of applications are provided, together with numerical experiments.

7: \end{abstract}

8: