Figure 1
Sample colonoscopy images from our database are shown as clear (a, b, c) and with motion artifact (d, e, f).
Figure 2
Original colonoscopy image (a), gray-scale representation of the original image (b), and the resultant image of adaptive histogram equalization applied on gray-scale image (c).
Table 1
Automatic clear vs. blurred image discrimination accuracies, specificities, f-measure scales, sensitivities and AUC for different feature extraction methods and classification approaches.
| Measurement Performance | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Classification Methods | ||||||||||||
| SVM | LDA | k-NN | ||||||||||
| LAP* | WT | DCT | All | LAP* | WAV | DCT | All | LAP* | WAV | DCT | All | |
| Accuracy | %76 | %71 | %66 | %85 | %72 | %67 | %69 | %85 | %72 | %64 | %63 | %70 |
| Specificity | %72 | %86 | %62 | %84 | %76 | %72 | %62 | %82 | %82 | %78 | %70 | %78 |
| f-measure | 0.76 | 0.61 | 0.65 | 0.85 | 0.72 | 0.66 | 0.68 | 0.85 | 0.70 | 0.61 | 0.52 | 0.69 |
| Sensitivity | %80 | %56 | %70 | %86 | %68 | %62 | %76 | %88 | %62 | %50 | %56 | %62 |
| AUC | 0.76 | 0.76 | 0.72 | 0.88 | 0.72 | 0.71 | 0.72 | 0.88 | 0.71 | 0.62 | 0.66 | 0.71 |
[i] * Refers proposed method for discrimination images with motion-artifact from clear images.