This article considers the problem of approximating a general asymptoticall
y smooth function in two variables, typically arising in integral formulati
ons of boundary value problems, by a sum of products of two functions in on
e variable. From these results an iterative algorithm for the low-rank appr
oximation of blocks of large unstructured matrices generated by asymptotica
lly smooth functions is developed. This algorithm uses only few entries fro
m the original block and since it has a natural stopping criterion the appr
oximative rank is not needed in advance.