A topological approach for detection of chessboard patterns for camera calibration - Emerging Trends in Image Processing, Computer Vision, and Pattern Recognition - page 443

FIGURE 6 Examples of Bouguet database images. (a) Image 1, (b) image 2, (c) image 4, (d)
image 9, (e) image 17, and (f) image 19.

Table 1
Results for the Bouguet Database

Image 1 2 3 4 5 6 7 8 9 10

Corners 151 154 147 153 51 149 132 110 143 153

% 96.79 98.72 94.23 98.08 32.69 95.51 84.62 70.51 91.67 98.08

Image 11 12 13 14 15 16 17 18 19 20

Corners 152 155 156 138 121 154 155 61 111 118

% 97.44 99.36 100.00 88.46 77.56 98.72 99.36 39.10 71.15 75.64

The rows named “Corners” list the number of detected x-corners and rows named “%” list the hit rate.

The mean accuracy of the algorithm is 85.38%; however, two images (images 5 and 18) re-
quire atention by the low percentage of detected points. They represent situations where the
calibration plane is very inclined to the camera. In this case it is expected that the corners are
uncharacterized by high perspective distortion and lack of focus in the image. Another as-
pect to be considered is that images in this database are of low contrast, which difficults the
identiication of alternations of high contrast. Figure 7 shows examples the best and worst for
Bouguet images.