Table 10 Data augmentation methods
From: Deep learning-driven multi-view multi-task image quality assessment method for chest CT image
| Â | Inspiration | Position | Radiation protection | Artifact |
|---|---|---|---|---|
Vertical flipping | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) |
Horizontal flipping | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) |
Random scaling | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) | Â |
Random cropping and padding | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) | Â |
Gaussian noise | \(\checkmark\) | Â | Â | Â |
Gaussian blur | \(\checkmark\) | Â | Â | Â |
Contrast limited adaptive histogram equalization | \(\checkmark\) | Â | \(\checkmark\) | \(\checkmark\) |
Edge-enhancing | \(\checkmark\) | Â | Â | \(\checkmark\) |
Brightness randomization | \(\checkmark\) | Â | \(\checkmark\) | \(\checkmark\) |
Perspective transformation | \(\checkmark\) | \(\checkmark\) | Â | Â |
Motion blur | \(\checkmark\) | Â | Â | Â |
Additional images | 2680 | 1234 | 690 | 2090 |