Precision distortion correction technique based on FOV model for wide-angle cameras in automotive sector - Emerging Trends in Image Processing, Computer Vision, and Pattern Recognition - page 457

In other words, it finds the minimum sum-of-squares difference between the actual straight-
line component and the projected line component. By using this function for distance com-
putation, the minimum distance to the straight-line component of row ( m ) and column ( n ) is
calculated to estimate the distortion center ( c x , c y ).
The distortion center estimation method using 2D paterns performs precise distortion
correction by finding the minimum distortion distance. Precise distortion correction can be
achieved by applying the least distortion distance estimation method to the FOV distortion
correction model using 2D paterns. An LUT is produced with regard to the distortion loca-
tions on the 2D plane using the estimated distortion center, thereby being applied to the actual
image.
4 Experiment and evaluation
In order to verify the result of the distortion correction algorithm, the following distortion cor-
rection experiment device was developed. The experiment environment was configured using
notebook computers, USB cameras, wide-angle lenses, and chessboard paterns as the hard-
ware coniguration. The target image was a 2D black and white chessboard patern with 6 ho-
rizontal rows and 11 vertical columns. One square of the chessboard was 163 pixel × 163 pixel while
the actual size was 45 mm 2 .
As shown in Figure 5 , the experimental environment was prepared to perform real-time
processing for the proposed algorithm, in which a 2D chessboard was positioned at the center
and a real-time camera image taken with a wide-angle lens was sent to the personal computer
for distortion correction.
FIGURE 5 Experiment environment for distortion correction using wide-angle lens camera.
Once the accurate distortion center (C) is set in the FOV model and the distortion is correc-
ted, all straight lines can be restored almost completely as shown in Figure 6 . This result is
obtained because the distortion center was set repeatedly in the distortion correction process.
FIGURE 6 Distortion correction using the FOV model.
However, additional distortion was generated as shown in Figure 7 when distortion cor-
rection was performed again assuming that the distortion center was the center point of the
image where there was an error from the lens distortion center. The analysis result of the dis-
tortion center showed that when a fine error of − 30 pixels in the X direction and + 30 pixels in
the Y direction was applied to the same image, a phenomenon representing a fine curve with
a specific directivity was found. This means that more severe distortion was found in propor-
tion to the distortion center. The reason for this fine center error while projecting it onto the