An approximate isometry with error is a map
between metric spaces
that preserves every distance to within additive error
,
It is also called an -isometry. Error zero gives an isometry.
For normed spaces, the distances
are induced by their vector norms (Hyers and Ulam
1945).
For real Hilbert spaces and
, suppose that
is an approximate isometry with
. There is a linear
transformation
that is an isometry such that, writing
for the orthogonal projection
onto the image of
,
for all
(Šemrl 2025). The map
is given by
. Turnšek (2026)
obtained the sharper perpendicular estimate
when , where
is the identity operator.
For every
, an explicit example from
to
attains equality at a point of vector
norm
. Consequently, the coefficient
in the corresponding general bound
is optimal. For
, the bound
holds.