Measurement grid:

The cross-correlation can be calculated with the pattern-box(3.2.1) centered at any point on the first image (a), yielding a corresponding velocity vector. These points are located on a given measurement grid, which is either regularly spaced (default option), or defined by a grid file of and coordinates. The velocity vectors correspond to the position at the middle between the initial pattern-box center and the optimum displaced one. For each point on the measurement grid(3.2.4), this point has coordinates where is the velocity vector expressed in pixel displacement. The velocity field is therefore never obtained on a regular grid. The velocity field can be obtained on a regular grid by interpolation(6.1), which also provides the spatial derivatives .

A typical mesh for this measurement grid is (for instance along ), so that adjacent correlation subboxs overlap by one half. A smaller mesh (e.g. ) leads to a 3/4 overlap. The resulting redundancy is useful to reduce the errors and evaluate them by continuity(). A velocity field on typically a 100x100 grid is then obtained.

Masks can be introduced to restrict the measurement domain. The goal is to restrict the processing to an illuminated sector, to mask the field beyond walls, or remove bad parts of the image. The mask is a 8bit image file (.raw or .png) with the same dimensions as the image to mask. A greyscale code is used: for an intensity <20, the velocity is set to 0 (for instance at a wall); for 20<intensity<200, the vector is undefined (but it will be determined by interpolation from neighbors in the patch(3.4) programme; for 200<intensity, the vector will be computed. Center of the pattern box?VERIFIER