4 Design and function
3DPIXA
CD40131 Version 1
25 / 99
Image point calculation:
For calculating the height of an object point P, the corresponding 2 image points P
L
and
P
R
have to be identified.
– Differences of the 2 raw images (e.g., orientation, optical resolution) are rectified by
calculation.
– A pattern matching approach is applied to the rectified images to find the
corresponding 2 image points (see chapter 4.4.1, p. 25).
The distance (disparity) of the 2 corresponding image points is measured with subpixel
accuracy. Each pair of corresponding image points results in a disparity value recorded in
the height map image. The height map image consists of 16 bit grey values.
The height in mm is determined from the height map image using the CS-3DAPI with
calibration and configuration file.
4.4.1
Pattern matching
For pattern matching a correlation-based approach is used.
Fig. 8:
Pattern matching
1
Front tab / Master image
2
Rear tab / Slave image
3
Reference pattern
4
Search window
5
Best match
6
Correlation values
For each image point in the first image (e.g. slave image) a small window surrounding this
point is taken as a reference pattern (3).
For this reference pattern the correlation values (6) are performed for all image points in the
expected height range (determined by the minimum and maximum disparity setting) using a
search window (4) in the second image (e.g. master image).
The highest correlation value gives the position of the best match (5). The position shift of the
2 image points is the disparity.
