The results from the first CIV processing (CIV1) can be improved in an iterative way by a succession of false vector removal (fix), interpolation to make a new guess, and CIV processing with reduced search zone, using the estimate from the previous processing. By this method convergence to local correlation maxima can be obtained, with displacements mutually consistent with their neighborhood.
Furthermore, in the second CIV processing (CIV2), the prior knowledge
of the local deformation rate and rotation is used to refine the pattern
matching, see fig.10. During the time lapse 
, the
local displacement differs from the pure translation by 
(or 
at the edge of the pattern-box. Deformation
will deeply affect CIV when this shift becomes of the same order as
the apparent particle size 
, or image correlation length.
Estimating 
from a typical velocity 
as 
,
we get that deformation effects are important for
.
Since 
pixel, with a displacement 
10
pixels, this corresponds to 
.
To describe such deformation, it is necessary to interpolate the image itself between the integer pixel values. This is done by a spline interpolation of the image intensity in the pattern-box.
Taking into account this deformation effect allows, as well as prior information, allows to increase the time interval, and therefore the precision. The use of prior information may also allow to reduce the pattern-box and increase the spatial resolution.
